Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
PRIORITY 6
SUSTAINABLE DEVELOPMENT
GLOBAL CHANGE & ECOSYSTEMS
INTEGRATED PROJECT
Sustainable Bridges SB-4.7 2007-11-30 2 (8)
This report is one of the deliverables from the Integrated Research Project Sustainable Bridges - Assessment for
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework Pro-
gramme. The Project aims to help European railways to meet increasing transportation demands, which can only
be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
passenger trains. This requires that the existing bridges within the network have to be upgraded without causing
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and econ-
omy of the railways.
A consortium, consisting of 32 partners drawn from railway bridge owners, consultants, contractors, research
institutes and universities, has carried out the Project, which has a gross budget of more than 10 million Euros.
The European Commission has provided substantial funding, with the balancing funding has been coming from
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule Tech-
nical University has undertaken the scientific leadership.
The Project has developed improved procedures and methods for inspection, testing, monitoring and condition
assessment, of railway bridges. Furthermore, it has developed advanced methodologies for assessing the safe
carrying capacity of bridges and better engineering solutions for repair and strengthening of bridges that are found
to be in need of attention.
The authors of this report have used their best endeavours to ensure that the information presented here is of the
highest quality. However, no liability can be accepted by the authors for any loss caused by its use.
Figure on the front page: Output from FEM analysis - principal stresses and displacement vector.
Summary
This background document contains the basic and full information summarized in chapter 8
of the Guideline SB-Resist (2007): Guideline for load and resistance assessment of railway
bridges, with reference to the assessment of masonry arch bridges.
Recent survey carried out by WP2 showed that 40% of existing rail network bridges
in Europe are masonry arches. More than 60% of those bridges are over 100 years
old and still carry ever increasing levels of loading and increasing volumes of traffic.
Due to the constantly increasing weight of rail traffic there is increasing demand for
better understanding of their life expectancy and fatigue limits. It is imperative that
the bridge stock is not adversely affected by these changes in the loading regime and
that appropriate assessment, modelling, repair and strengthening techniques are
available.
Over the past 10 years there has been an extensive programme of research which
considered some aspects of masonry arch behaviour. A number of small and large
scale tests have been carried out on masonry arches, most of them have been under
static (monotonic) loading and considered mainly the arch ring itself. There has been
also very little work done on investigating the long-term effect of traffic (cyclic) load-
ing and the effect of deteriorated masonry on the fatigue life of the bridge.
Most tests focussed on the behaviour of the arch ring itself and little work has been
done to investigate how the loads from the train traffic are transferred onto the arch
ring and how much the soil-structure interaction may contribute towards the capacity
of the bridge.
Additionally, most of the available methods of assessment are deterministic. The
proper treatment of uncertainty in the assessment of metal or concrete bridge can
save it from strengthening or replacement. Many practical cases have shown the po-
tentiality of reliability-based assessment methods for concrete and steel bridges.
However, their application to masonry bridges has been almost negligible, despite
some experiences have shown the potentiality of the use of such methods in the
case of masonry arch bridges.
The background document is an attempt to address and solve the above questions.
To this end, the document is divided in the following parts:
Part D4.7.1 includes the analysis of failure mechanism in multi-ring brickwork arches
as well as the fatigue performance of multi-ring brickwork arches and the soil-
structure interaction.
A series of tests have been carried out at the University of Salford on large-scale
multi-ring brickwork arches (3m and 5m span, two and three-ring) under long term
Sustainable Bridges SB-4.7 2007-11-30 4 (8)
cyclic loading. Tests showed that although the classical mode of failure of arches un-
der static loading is generally seen as the four-hinge-mechanism, all arches within
the present test series failed by ring separation under fatigue loading. The fatigue
loading reduced the load capacity to as low as 37% of the static load capacity for
two-ring arches and 57% for three-ring arches and was a function of ring separation.
Old masonry arch bridges often have suffered serious material deterioration as a
consequence of environmental effects, cyclic/fatigue effect, water penetration, etc
which can seriously reduce the load bearing capacity of the bridge and change its
critical modes of failure. Of particular concern is mortar loss from between the rings
of a multi-ring brickwork arch, resulting in ring separation with the consequential loss
of carrying capacity. A series of 5m span arches built of weak bricks were tested to
represent aged, poor quality masonry. Weak brick arches which have been tested so
far have shown reduced fatigue capacity of the arch by ca. 20% but did not greatly
change the mode of failure compared to strong brick arches.
Based on the test results, a model for an interactive S-N (ISN) curve has been sug-
gested. The ISN curve could be developed for each bridge, based upon standard
curves and modified according to the engineers assessment of the structural condi-
tion. It would allow qualitative assessment of the residual life and critical modes of
failures. It would also help decide the appropriate repair technique and quantify the
new residual life for a given load regime. The anticipated ISN curve based on the test
results is also presented in the study and the effect of material deterioration have
also been included.
Results from three large-scale bridge tests have demonstrated the importance of the
type of backfill and support conditions and have produced high quality data that can
be used to validate numerical models. They also indicated that the test rig performed
as designed (giving effectively plane strain conditions); inclusion of large windows
along one side of the test chamber permitted the acquisition of good particle image
velocimetry (PIV) data which is being used to help better understand the nature of
the soil-arch interaction.
Part D4.7.2 deals with the soil-structure interaction and the dynamic response of em-
bankment and backfill to railway loading.
In order to model and assess the load capacity of arch bridges accurately, it is nec-
essary to gain clear understanding of how much of the axle loads pass through the
backfill and how the arch is seeing the passing traffic. The influence of higher axle
loads and higher train speed on the properties/behaviour of the backfill and em-
bankment material will be investigated by numerical modeling. The long-term effects
of increased train loads and increased train speed on the stress limits of the structure
will be investigated in order to model the residual life of the bridge.
The SGI has studied the dynamic soil-loading problem using typical embankment
parameters and representing the bridge structures by rigid boundaries. The numeri-
cal modelling technique FLAC3D was used. The scope of this work included perform-
ing numerical analyses to evaluate the load distribution and deflections in bridge
transition zones due to the passing of trains. No field or laboratory tests were per-
formed. Numerical analyses consisted of three-dimensional, dynamic analyses. Us-
ing a bridge representative of a typical European concrete bridge, a numerical para-
Sustainable Bridges SB-4.7 2007-11-30 5 (8)
metric study was performed to study the effects of 1.) the train load, 2.) the train ve-
locity, and 3.) the stiffness of the ballast, sub-ballast and the embankment fill. Key
findings from the work include:
The greatest increases in vertical and horizontal stresses are concentrated
within the upper 1.0 to 1.2m of the system. The additional vertical and horizon-
tal stresses caused by the passing train decrease with depth.
As train axle loads increase, the vertical and horizontal stresses, and vertical
deflections beneath the train increase.
As train velocity increases, the horizontal stresses and vertical deflections in-
crease, however, they do not increase linearly with velocity. As the velocity of
the train approaches the Rayleigh wave velocity of the embankment system,
significant horizontal stresses and vertical deflections develop.
Vertical pressures are affected more by train axle load, and the horizontal
pressures by train velocity.
As the stiffness of the ballast and/or sub-ballast materials increase, net hori-
zontal stresses exerted on the bridge decrease. Horizontal stresses in the
embankment fill are relatively small, but increase when a soft fill is used.
When a soft embankment fill is used, vertical deflections increase substan-
tially, and these deflections are observed to occur 3m behind the rain, as if
there is a stern wave behind the train.
Numerical analyses were also performed using a bridge geometry representative of
European masonry arch bridges. It can be assumed that the factors that affect the
stress distribution and deflections for the concrete bridge structure (as described
above) will have the same effect for the masonry arch bridge structure. However, the
geometry of the arch bridge is different than the concrete bridge in two significant
ways: (1) for masonry arch bridges there is a gradual change in stiffness, while for a
concrete bridge the change is more abrupt, and (2) for the masonry arch structure,
the ballast and subballast are confined within a relatively small volume. Due to the
latter factor, additional vertical and horizontal stresses induced by the train are dis-
tributed immediately to the arch structure. Since there is less soil above the crown of
the arch to dissipate stresses, the crown of the arch will experience the greatest in-
crease in vertical stress under the passing of a train.
The above work was verified against data from an instrumented test embankment
published by the Finnish Railway administration. These data referred to an embank-
ment which did not include a bridge. The results give some insight into the behaviour
in bridge transition zones. Due to the lack of funding, field tests and/or laboratory
tests in a real bridge close to an embankment were not developed to verify the nu-
merical analyses.
them are suitable for modelling all types of existing damages in the bridge. The pre-
sent part of the background document gives an overview of the masonry arch dam-
ages and guidance on which assessment methods are suitable for various types of
damages and on their limitations for use.
Finally, part D4.7.4 reports a feasibility study of the application of probabilistic tech-
niques to the assessment of existing masonry arches.
Probabilistic assessment methods are not generally used for masonry arch bridges.
One of the reasons of its limited application is the difficulty of defining reliable failure
criteria for these types of structures. Experimental tests have shown that failure of an
arch is generally of a global nature rather than by failure of individual components.
There is also a lack of accurate theoretical models to idealise the behaviour of ma-
sonry arch bridges as a system.
Part D4.7.4 of the background document presents a complete methodology for the
probabilistic assessment of masonry arches at the serviceability and ultimate limit
states. The document explains the definition of the different failure modes and corre-
sponding limit state functions that may occur depending on the type of masonry con-
struction (mainly single-ring and multi-ring). Also the introduction of the possibility of
the fatigue failure of masonry arch bridges and the proposal of new assessment
methods based at the serviceability level and not only at the ultimate level of the 4
hinge-mechanism is a promising initiative in the field. As a consequence, the present
document has deeply investigated the potentiality of a probabilistic approach in the
estimation of fatigue safety and remaining service life of masonry arch bridges. As a
result, the need of simple numerical models to perform fast simulation trials has been
enhanced. Thus, the influence of important features in the bridge response, as the
arch-backfill interaction, the modelling of multi-ring arches and the bridge skewness
in the development of simplified theoretical models at the ultimate and serviceability
limit states have been worked out too.
The analysis developed in D4.7.4 has verified the results obtained in laboratory tests
regarding the influence of skewness in the failure load by 4-hinge mechanism. In fact,
it has been confirmed that the influence of a skew angle less than 22.5 is almost
negligible. The study has confirmed that only with advanced 3D FEM models is pos-
sible to correctly model the behaviour up to failure of masonry arches with skew big-
ger than 22.5. It has been also concluded that the worst load position for the arch
can be deduced using a simple model.
In summary, a methodology is presented for the fatigue and serviceability assess-
ment of masonry arches, and probabilistic S-N curves for the behaviour of masonry
under fatigue are developed based on the few experimental data available. In the
case of masonry under compression, a fatigue equation with various levels of prob-
ability of failure is also proposed that may be used for deterministic assessments. In
the case of masonry under shear a fatigue equation with a 50 % confidence level is
also proposed. The last can be of relevant application to the case of multi-ring
arches. Of course, these are only preliminary models and these fatigue models have
to be updated as more experimental data becomes available from future laboratory
and full-scale tests. Finally, the examples presented show how a probabilistic as-
sessment of masonry arch bridges which were rated unsafe using standard linear
Sustainable Bridges SB-4.7 2007-11-30 7 (8)
models, indicated sufficiently high level of safety margins when assessed together
with appropriate non-linear models.
In the course of the research it became apparent that the current methods of as-
sessment lack a single multi-level (initial, intermediate, enhanced) approach that in-
cludes all the current methods of assessment/analysis and gives clear guidance on
the philosophy that governs the determination of the safe working loads and ultimate
load carrying capacity.
The SMART (Sustainable Masonry Arch Resistance Technique) assessment
method, developed as a result of the present research, achieves this. There are 7
steps in the method which are summarized in the following table:
Statically determi-
nate analysis to
establish stresses
5) ULS N/A Check that the Check that the
failure load is failure load is
greater than the greater than the
factored loading factored loading
6)PLS Determined by the Check that the Check that the
application of fac- working loads do working loads do
tors to the initial not induce not induce
permitted axle stresses greater stresses that are
loading than permissible greater than the
stresses permissible
stresses if the
do then probabilis-
Sustainable Bridges SB-4.7 2007-11-30 8 (8)
The Ultimate Limit State (ULS) is defined as the condition at which a collapse
mechanism forms in the structure or its supports. The Permissible Limit State (PLS)
is defined as the limit at which there is a loss of structural integrity which will meas-
urably affect the ability of the bridge to carry its working loads for the expected life of
the bridge.
The writing of the different parts as well as the review has been carried out according
to what is presented in the following table:
PRIORITY 6
SUSTAINABLE DEVELOPMENT
GLOBAL CHANGE & ECOSYSTEMS
INTEGRATED PROJECT
Sustainable Bridges SB-4.7.1 2007-11-30 2 (72)
This report is one of the deliverables from the Integrated Research Project Sustainable Bridges - Assessment for
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework
Programme. The Project aims to help European railways to meet increasing transportation demands, which can
only be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
passenger trains. This requires that the existing bridges within the network have to be upgraded without causing
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and
economy of the railways.
A consortium, consisting of 32 partners drawn from railway bridge owners, consultants, contractors, research
institutes and universities, has carried out the Project, which has a gross budget of more than 10 million Euros.
The European Commission has provided substantial funding, with the balancing funding has been coming from
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule
Technical University has undertaken the scientific leadership.
The Project has developed improved procedures and methods for inspection, testing, monitoring and condition
assessment, of railway bridges. Furthermore, it has developed advanced methodologies for assessing the safe
carrying capacity of bridges and better engineering solutions for repair and strengthening of bridges that are found
to be in need of attention.
The authors of this report have used their best endeavours to ensure that the information presented here is of the
highest quality. However, no liability can be accepted by the authors for any loss caused by its use.
Table of Contents
Summary .................................................................................................................... 5
1. Masonry arch behaviour under long-term cyclic loading ........................................ 7
1.1 Background.................................................................................................... 7
1.2 Objectives ...................................................................................................... 8
1.3 Arch construction ........................................................................................... 8
1.3.1 Arch dimensions ................................................................................. 9
1.3.2 Material properties .............................................................................. 9
1.3.3 Loading arrangement........................................................................ 10
1.3.4 Instrumentation ................................................................................. 11
1.4 Static loading tests....................................................................................... 12
1.4.1 Strong brick arches ........................................................................... 12
1.4.2 Weak brick arches ............................................................................ 19
1.5 Cyclic loading tests ...................................................................................... 21
1.5.1 Strong brick arches ........................................................................... 21
1.5.2 Weak brick arches ............................................................................ 28
1.6 Summary ..................................................................................................... 32
1.6.1 Static loading summary..................................................................... 32
1.6.2 Cyclic loading summary .................................................................... 34
1.6.3 SN curves ......................................................................................... 35
2. Soil-structure interaction in masonry arch bridges................................................ 37
2.1 Background.................................................................................................. 37
2.2 Objectives .................................................................................................... 37
2.3 Experimental programme............................................................................. 38
2.3.1 Test rig.............................................................................................. 38
2.3.2 Bridge geometry and material properties .......................................... 39
2.3.3 Bridge construction ........................................................................... 41
2.3.4 Bridge instrumentation ...................................................................... 45
2.3.5 Test Procedure and loading arrangement......................................... 48
2.3.6 Test results ....................................................................................... 49
2.4 Finite element modelling .............................................................................. 54
2.4.1 Finite element modelling approach ................................................... 54
2.4.2 Case study - flexible and smooth strip footings on stratum of clay.... 58
Sustainable Bridges SB-4.7.1 2007-11-30 4 (72)
Summary
The Sustainable Bridges project survey has shown that 40% of the existing rail
network bridges in Europe are masonry arch bridges and that more than 60% of them
are over 100 years old. It is imperative that the bridge stock is not adversely affected
by changes in the loading regime and bridge condition and that appropriate
assessment, modelling, repair and strengthening techniques are available.
Research at the University of Salford has focused on two main issues relating to
masonry arch bridges that can significantly influence the bridges capacity and
service life:
fatigue performance of multi-ring brickwork arches
soil-structure interaction in masonry arch bridges.
Research described within the current project and by others has confirmed that
masonry arch bridges behave in a complex way and that simple semi-empirical
assessment methods are no longer appropriate to justify an increase in load carrying
capacity. Consequently, a new holistic masonry arch bridge assessment strategy
(Sustainable Masonry Arch Resistance Technique - SMART) has been proposed that
incorporates the currently available assessment methods and takes recent research
Sustainable Bridges SB-4.7.1 2007-11-30 6 (72)
and findings into account. The new Sustainable Masonry Arch Resistance Technique
(SMART) sets out a strategy by which the residual life and permissible limit of
masonry arch bridges can eventually be determined with the help of a set of SN
curves (see D4.2 Chapter 8 for details). It can help prioritise conflicting maintenance
demands on limited budgets and improve asset management. The SMART
assessment method is currently in its infantry and large volumes of material test data
are required for further development of the technique.
Sustainable Bridges SB-4.7.1 2007-11-30 7 (72)
Railway bridges in Europe almost invariably contain multiple arch rings. These rings
are either simply connected by uninterrupted mortar joints or crossed by headers
which provide an interlocking effect and greatly increases the shear strength of the
connection between rings (see Figure 1.2). The type of arch construction shows
certain regional consistencies around Europe, for example arch bridges in Britain are
generally built without headers, while in Southern Europe arches with interlocking
headers are more common.
The shear resistance of arch rings is also influenced by the quality of mortar and
brick units. Aged arch bridges often suffer from weak shear connections from washed
out mortar joints and/or weak interlocking bricks that can result in unexpected ring
separation and may significantly reduce the load carrying capacity of the bridge.
Sustainable Bridges SB-4.7.1 2007-11-30 8 (72)
The currently described test series was designed to represent multi-ring arches
without headers under long-term traffic loading to investigate their likely modes of
failure and fatigue capacity.
1.2 Objectives
Objectives of the test series were to investigate the behaviour, possible modes of
failure, capacity and permissible limit state of multi-ring masonry arches under long-
term fatigue loading.
Hydraulic jack
1.3.4 Instrumentation
Horizontal and vertical deflection was measured at the and span using LDVTs
(see Figure 1.5).
H () H ()
V () V ()
Vibrating wire gauges (GAUGE TYPE T/S/R, 139mm gauge length) were used to
monitor surface strain with a range of 0.5 - 3000 microstrains and were attached to
the masonry surface with an epoxy resin adhesive.
In addition to the standard test methods acoustic emission (AE) technique was
applied during the tests to gain deeper understanding of the fracture process. For
acoustic emission (AE) monitoring a Physical Acoustics DiSP system was used with
four channel PCI DSP4 boards and filter set to 20-200kHz. Preamplifiers were IL40D
with 20-100kHz frequency. The sensors were Physical Acoustics R6D resonant
piezoelectric sensors with a response range of 40-100kHz. The sensors were
selected based on advice and past experience as a good compromise of detection /
distance and avoiding noise from the loading rig. The filter bandwidth was broader
than the preamplifier and sensors to ensure that signals were not distorted. AE WIN
software was used to process data and in addition to the AE hit recordings load and
deflection were also measured every time emission was received. Up to eight AE
sensors were attached to the brick surface using a thin layer of hot-melt glue. This
technique had been tested to provide good coupling for transmitting AE signals.
Sustainable Bridges SB-4.7.1 2007-11-30 12 (72)
DL + LL DL
23kN
6.5kN
14kN (?)
Horizontal and vertical deflection was measured under the and span. No
changes in deflection were recorded during the opening of the 1st and 2nd hinges (see
in Figure 1.8). First change in deflection was shown from around 21kN, just before
the 3rd hinge was observed. Deflection sensors were removed from the arch at 23kN
to avoid damage.
DL + LL DL
AE 1
AE3
22kN 28kN
AE5
AE6
AE4
(?) (?)
Surface strain was measured in 5 sections along the arch and 4 locations in each
section as shown in Figure 1.12. Strain measurements against the load at the
span and span is shown in Figure 1.13. Clear change in strain at the span by
gauges B, C and D form 14kN load indicates a radial hinge opening from the intrados
way through the arch section. No change was recorded by strain gauge A at the
top of the section. Changes in strain and deflection (Figure 1.11) around 14kN load
indicate the opening of the first hinge, however the first hinge was visually observed
only around 22kN.
The final change in strain is shown at 27kN which again agrees well with the changes
in deflection. The fourth hinge is therefore likely to have opened at 27kN which
caused the arch to become unstable at 287kN.
Strain measurements have therefore provided useful information regarding changes
in behaviour and opening of hinges.
A E
B F
C G
D H
200
0
Microstrain
-200
-400 A
B
-600 C
D
-800
0
10
12
14
16
18
20
22
24
26
28
Load (kN)
200
E
150 F
G
100 H
Microstrain
50
-50
-100
0
10
12
14
16
18
20
22
24
26
28
Load (kN)
Figure 1.13: Arch G: Load-Strain
During load application 6 acoustic emission sensors were additionally attached to the
arch surface. For AE sensor location see Figure 1.9. Recorded emission is shown in
Figure 1.14 for amplitude vs. load. Unfortunately the amplitude limits were set
reasonable high (55dB), hence no information was recorded for low-level emission.
The first significant emission was recorded by AE1 at the span at 14kN. At 22-
23kN AE3 indicated large emission and associated radial crack opening around the
span which only became noticeable with the naked eye around 28kN when the
arch failed.
Amplitude (dB) Amplitude (dB) Amplitude (dB)
100 100
AE1 AE4
80 80
60 60
40 40
20 20
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
100 100
AE2 AE5
80 80
60 60
40 40
20 20
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
100 100
AE3 AE6
80 80
60 60
40 40
20 20
0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30
DL + LL DL
AE3
AE2
AE4
AE1
30kN
DL
AE6
AE7
AE5
AE8
8 acoustic emission sensors were attached to the arch surface during load
application (see
Figure 1.15 for gauge locations AE1-AE8 and Figure 1.19 for AE amplitude vs. load
recordings). Although no visual signs of damage could be seen until the arch failed
by sudden ring separation under 30kN, AE recordings showed intense activity by
sensors AE1 and AE5 from the start of live load application. Location of intense AE
activity coincided exactly with the location of final ring separation and identified the
Amplitude (dB) Amplitude (dB) Amplitude (dB) Amplitude (dB)
DL + LL DL
31kN 30kN
28kN
26kN
AE4
AE5
AE3
AE6
AE2
AE7
30kN 14kN
AE8
7 acoustic emission gauges were attached to the arch intrados (for AE gauge
locations see Figure 1.20). AE amplitude vs. load recordings showed increased
activity form the start of load application by sensors AE2 and AE3 which clearly
identified the radial crack opening at 14kN in their vicinity. Sensors AE2, AE3 and
AE4 also show increased activity from ca. 22kN, well in advance of ring separation
failure at 26kN load.
14kN
14kN (23000 c.)
(23000 c.)
9kN 14kN
9kN 11kN
(23000 c.)
(c. = cycles)
Sliding
12kN
(23000 c.)
DL
DL DL
8kN DL
Drying cracks
(0kN)
0kN
? ?
8 acoustic emission sensors were attached to the arch intrados during the loading
series (for AE sensor locations see Figure 1.31). The average released absolute
energy per cycle is summarised for the beginning and end of each 106 cycles for
each load level (i.e. for 0 cycle and 106 cycles) in Figure 1.35. The released energy
Sustainable Bridges SB-4.7.1 2007-11-30 27 (72)
level was relatively low for 10kN and 12kN and started to increase from the start of
14kN load application. Under 18kN the released energy increased further until failure
occurred. With the help of acoustic emission recordings the 14kN load level could be
identified as the fatigue (permissible) limit for the arch, above which accelerated
damage propagation took place and caused eventual failure.
Abs. Energy (aJ) / Cycle
120
100
Permissible
separation
80
Ring
limit
60
40
20
0
10kN
10kN
12kN
12kN
14kN
14kN
16kN
16kN
18kN
18kN
Loading history
Figure 1.35: Arch O: AE Absolute energy per Cycle vs. Loading history
Sustainable Bridges SB-4.7.1 2007-11-30 28 (72)
0kN
25.5kN 13.5kN
25.5kN
0kN 16.5kN
(680c.)
1.6 Summary
DL + LL DL DL + LL DL
30kN
DL
14kN
Despite the different failure mechanisms, average static capacity was very similar for
3m and 5m span arches: 28.5kN and 28kN for the 3m and 5m arches respectively.
Capacity of the 3m arches was determined by the flexural tensile strength of masonry
that caused formation of a four-hinge mechanism, while capacity of the 5m arches
was determined by the shear strength of masonry and caused ring separation failure.
Although deflection can easily be measured on arch bridges, measurements have
only shown changes in the archs behaviour well after the development of significant
cracks. Deflection measurements may therefore not be reliable indications of the
archs condition. Strain measurements were able to identify crack development
immediately after their occurrence but only when measured in the appropriate arch
sections. Acoustic emission (AE) technique was able to indicate internal damage
propagation prior to crack opening and accurately identify future areas of damage. It
showed great potentials for condition assessment of masonry arch bridges.
Sustainable Bridges SB-4.7.1 2007-11-30 34 (72)
Initial loading
c) Arch O (Strong)
12kN (20,000c.) 12kN (2,400c.) 16.5kN
(680c.)
1.6.3 SN curves
The summary of the arch test series is shown in Table 1-13. Arches generally
developed a number of radial cracks relatively early during load application that could
be used for identifying likely modes of failure. Figure 1.45 shows a diagrammic
summary of visual defects with their likely modes of failure from the previously
described tests.
Table 1-13. Arch test summary
% of average
Loading
A 29 100% 1 Four-hinge
3m
G Strong 28 (28.5kN) 1 Four-hinge
Static
M 30 1 Ring separation
5m 100% (28kN)
T Weak 26 1 Ring separation
C 14 50 23,000 Ring separation
3m
E Strong 12 43 23,000 Ring separation
Cyclic
2 radial cracks ,
Ring separation
3 radial cracks 0, , 1
4-hinge mechanism
0, , 1
Abutment movement
4 radial cracks 0, , , 1
Ring separation
0, , , 1
4-hinge mechanism
0, , , 1
Sliding
Under static loading both 3m and 5m span arches failed at similar load levels but by
different mechanisms. Under long-term fatigue loading the permissible limit for the
arch series was around 40% of the static capacity, above which the arches failed
within a relatively low number of cycles.
A possible way of relating load levels to the number of cycles an arch can carry
before failure is with the help of SN curves. Arches can fail by a number of
mechanisms, e.g. four hinge mechanism, ring separation, sliding, etc., all of which
should be represented by individual SN curves. Figure 1.46 shows an indication of
SN curves for ring separation failure from the previously described arch tests series.
Stresses are expressed as percentage of the maximum static stress against and are
related to the number of cycles at failure. The two different brick qualities (Class A
engineering and Britley Olde English bricks) used in the test series are graphically
identified and bands for different brickwork qualities (A, B, C) are suggested. Large
volumes of further test data are now needed to develop SN curves for various failure
modes and masonry qualities.
Subsequently the SN curves can be incorporated into the newly developed SMART
assessment technique to provide a holistic assessment tool that can identify the
bridges service life under any working stress level and their likely modes of failure.
The SN curves can also be used to indicate the possible sensitivity of the bridges life
expectancy to changes in working stresses levels. If for example the shear stresses
increase from 50 to 55% of the max. static stress, life expectancy of the bridge may
reduce almost to its quarter (from 300,000 to 80,000 cycles) as shown in Figure 1.46.
SN curves can also indicate the permissible limits for each failure mode and material
quality below which no residual damage is likely to occur to the arch and loading can
be applied for a theoretically unlimited number of cycles without requiring
reinforcement or strengthening works.
100
Class A
Engineering bricks
Max. static stress (%)
80
Britley Olde
English bricks
60
55
50 Permissible limit
40
A
20 B
0
80,000
300,000
1,E+00
1,E+02
1,E+04
1,E+06
1,E+08
1,E+10
1,E+12
No. cycles
2.2 Objectives
Three bridges have been tested to failure to date, with the following objectives:
To test out the performance of the apparatus (tank, imaging and
instrumentation).
To determine the influence of abutment fixity on arch behaviour.
To provide a benchmark test result.
To better understand the behaviour of masonry arch bridges with clay backfill.
The first test bridge was designed to be similar to the 3m span bridges tested at
Bolton in the 1990s (Melbourne and Gilbert 1995), thereby permitting direct
comparison. However, unlike the Bolton bridges, which had been constructed
between rigid abutments potentially movable abutments were specified and
furthermore the walls of the plane-strain testing tank marked the edges of the bridge,
rather than the brickwork spandrel walls used previously.
The second test bridge was designed to be identical to the first with the exception
that fill material below the level of the crown of the arch was replaced with soft clay,
representative of that found in some local authority owned bridges in the UK.
Crushed limestone was used above the crown to reflect the fact that competent near
surface road/sub-base material is normally present in real bridges. Additionally, had
Sustainable Bridges SB-4.7.1 2007-11-30 38 (72)
this not been used then a local failure of the soil in the vicinity of the applied load
would have been likely.
The third bridge was successfully tested on 14 June 2007 and the test results are
currently being processed and should be completed before the end of the contract
(results, where available, are included in this document). The third test bridge was
filled with the same limestone used for the first bridge test. Unlike Bridge 1, twelve
pressure cells were incorporated in this bridge, enabling valuable additional
information to be obtained.
Plan view
Elevation
215
305
1665
750
395
550 180 3373 550
8300
elevation
215
305
1665
750
395
8300
elevation
750
395
8300
total of 10 shear box tests were carried out on the material for Bridge1. The average
density and strength are presented in Table 2-1. The shear properties did display
stress dependence. Dilation is clearly suppressed at higher stresses.
The second bridge (Bridge2) was backfilled with clay to crown level with 305mm of
compacted crushed limestone above. The clay was supplied by Marchington Stone
Ltd and was described as a firm red-dish brown slightly sandy CLAY with occasional
gravel. The suppliers determined index properties were as follows: natural moisture
content 15%; optimum moisture content 9%; Liguid Limit 29%; Plastic Limit 12%. The
average moisture content on material placement was 13.3%. The average density
and shear strength from undrained tests on eight samples taken after the test are
presented in Table 2-1.
Table 2-1. Material properties
Material Type
Masonry units Compressive strength (N/mm2) 154
Unit weight (kN/ m3) 23.2
Mortar Compressive strength (N/mm2) 1.9
Unit weight (kN/ m3) 14.4 15.4
Masonry Compressive strength (N/mm2) 24.5
Unit weight (kN/ m3) 21.6
Backfill Unit weight (kN/ m3) Limestone: 19.1
Clay: 22.1
(degrees) Limestone: 46.4
2
Cohesion (kN/mm ) Limestone: 22.4
Clay: 78
2.3.3.1 Abutments
Two reinforced concrete abutments on which the arch barrel were built were fixed
3000 mm apart, parallel to each other. The dimensions of the two abutments are
shown in Figure 2.6.
Each abutment comprised two parts; a lower section that was bolted to the structural
strong floor, and an upper section that could slide should the forces be large enough
to overcome the bond/friction between the blocks.
Sustainable Bridges SB-4.7.1 2007-11-30 42 (72)
230 154
215
13
530
530
20
395
322 325
2.3.3.2 Centering
The arch was constructed on custom made steel centering (as shown in Figure 2.7)
on which 101.6mm x 50.8mm x 1000mm planks were placed, in turn covered by a
sheet of plastic in order to minimise bonding of the masonry to the planks and
facilitate easy removal of the centring. Each curved steel beam was supported by two
individual stacks of bricks.
215
crown
750
3000
2.3.3.4 Tank
Following construction of the arch, the test tank was assembled around the arch. The
average width of the test rig was 1045mm while the average width of the arch was
1010mm. Tie bars were placed across the top and bottom of the frame to provide
requisite lateral stiffness.
In order to prevent fill falling between the gap between the arch and the test rig walls,
strips of closed cell foam were hot glued along the edges of the arch extrados so as
to span the gap (see Figure 2.8). The flexible foam would accommodate any minor
lateral arch movements while retaining the fill above.
Sustainable Bridges SB-4.7.1 2007-11-30 43 (72)
Figure 2.8: Gap covered with foam and attached latex marked
2.3.3.6 Backfill
The fills were compacted in layers using vibrating compaction plates. Sensitive areas
adjacent to the walls and the arches were compacted with a hand rammer.
For the fill comprised MOT Type 1 graded crushed limestone (Bridge1 and Bridge3)
the required weight for each layer was loaded into a crane mounted hopper for
transfer to the rig and spread evenly using a shovel to the required thickness, and
then compacted to the required specification using 10kN (1t) whacker plate.
The use of greased latex sheeting unfortunately reduced the visibility of the fill
material slightly. In order to ensure the images captured had sufficient contrast to be
processed using the GeoPIV software, 50mm gaps on the inside of both walls were
set up. The gaps between the plates and the walls were filled with coloured fill, as
shown in Figure 2.9, while normal fill was placed in the large central gap between the
plates. Figure 2.10 shows the completion of the filling of limestone for Bridge1 and
Bridge3.
Sustainable Bridges SB-4.7.1 2007-11-30 44 (72)
Figure 2.9: Gaps set up and filled with coloured fill to enhance the post test image
analysis
Bridge1 Bridge3
Filling and compacting of clay on such a large scale is challenging. Trial compactions
were carried out involving hand compaction in a compaction mould (Proctor
Compaction Test) and compaction on small scale 1m2 box using a vibrotamper at the
north end of the fully assembled tank. Guided by the experience obtained from these
tests, a filling methodology was established. The clay was first wetted up in batches
as appropriate to the required consistency and thoroughly mixed using the back-
acting arm of an excavator. It was then transferred to the test rig using the excavator
bucket and spread evenly using a shovel to the required thickness. Each layer was
then compacted using a 10kN vibrating compaction plate suspended from a crane to
ease handling. Sensitive areas adjacent to the walls and arch itself were compacted
with a hand rammer. Each pressure cell was hand covered using clay. No
compaction was employed directly adjacent or over the cells.
During clay filling, small soil samples were taken at regular intervals for moisture
content testing. A total of 95 samples were collected and the average moisture
content was 13.4%. Readings from a pocket penetrometer were also taken at regular
distance after compaction of each layer during the whole filling process.
To protect the clay from drying out after compaction into the tank, several cloth and
plastic sheets were used to cover the clay during and after filling. Figure 2.11 shows
the procedures of clay filling and Figure 2.12 shows the completion of the filling of
clay.
Sustainable Bridges SB-4.7.1 2007-11-30 45 (72)
The Bridge2 filling was finished with two layers of limestone on the top of the clay.
2.3.4.1 Deflection
Linear Variable Differential Transformer (LVDT) type displacement transducers were
placed beneath the intrados of the arch barrel and the top parts of the abutments to
measure its movement. The accuracy of these gauges was about 0.01mm.
Additionally, LVDT and/or mechanical dial gauges were being used to monitor lateral
movements of the test rig, to ensure that the conditions of plane strain are met.
Figure 2.13 and Figure 2.14 show the positions of these gauges for Bridge1 and
Bridge3 respectively.
Sustainable Bridges SB-4.7.1 2007-11-30 46 (72)
Figure 2.13: Layout of displacement transducers monitoring the arch and the test rig
(Bridge1)
G-3/4L G-1/4L
West East
abutment G-West G-East abutment
LV9 - Top LV12 - Top
LV10 - Middle LV13 - Middle
LV11 - Bottom LV14 - Bottom
West abutment
219
East abutment
750
920
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Figure 2.14: Layout of displacement transducers monitoring the arch and the test rig
(Bridge3)
PC12 PC11 PC10 PC9 PC8 PC7 PC6 PC5 PC4 PC3 PC2 PC1
West abutment
East abutment
Figure 2.15: Positions of pressure cells on arch extrados (Bridge3)
2.3.4.3 Imaging
The testing tank provides 14 bays between the steel columns along its length. Of
these, the middle 12 bays incorporate acrylic windows. In order to capture soil
displacements, a set of six SONY DSC-V1, 5 MegaPixel, digital cameras are set up
such that each camera images a pair of bays (as shown in Figure 2.16). During
testing, and following the application of each load increment, the cameras can be
remotely triggered in quick succession to capture the images. Illumination from above
each bay with a halogen lamp has proved essential to ensure good image quality and
to minimise reflections.
South North
abutment abutment
Figure 2.17: Layout of Acoustic gauges (Bridge1)
Sustainable Bridges SB-4.7.1 2007-11-30 48 (72)
loading beam
arch barrel
Section A-A
A
north abutment
south abutment
219
1045
wood base
920
750
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
A
The third limestone filled bridge (Bridge3), which was tested on 14 June 2007, was
carried out in the newly relocated laboratory. Figure 2.19 shows the similar loading
arrangement for Bridge3 test.
Sustainable Bridges SB-4.7.1 2007-11-30 49 (72)
South North
abutment abutment
South North
abutment abutment
West East
abutment abutment
The limestone filled (Bridge1) failed at a load of approximately 125kN, the clay filled
bridge (Bridge2) failed at a load of approximately 90kN, and recently tested Bridge 3
failed at the load of approximately 148kN.
Figure 2.21 shows the bridge and surrounding limestone backfill close to the point of
collapse from the observation of the large windows.
Sustainable Bridges SB-4.7.1 2007-11-30 51 (72)
Cracks, especially at/near both abutments, were clearly seen after one side of the
walls was taken away. Figure 2.22 shows the cracks in the arch barrel after Bridge3
test. Figure 2.23 shows the standing bridge after one side of the wall was removed.
160
140
120
80
Bridge2 (clay filled)
60
40
20
0
0 5 10 15 20 25 30
Radial displacement (mm)
Figure 2.25: Load Images of limestone filled bridge with PIV soil displacement
vectors superimposed
Figure 2.26: Images of clay filled bridge with PIV soil displacement vectors
superimposed
computational cost; and secondly, correct material properties and constitutive laws
for the units and mortar joints are difficult to define.
In the level of the simplified micro-modelling, masonry is represented by continuum
elements where the behaviour of the mortar and unit-mortar interfaces is combined to
give a discontinuous element, as shown in Figure 2.28 (b); with this approach, it is
possible to consider the masonry as a set of elastic blocks bonded together by
potential fracture lines at the joints. In this level of modelling, only one interface
element is used for the mortar joint, including both areas of adhesion, so that
considerably fewer elements are necessary compared with detailed micro-modelling.
A disadvantage here is that Poissons effect of mortar is ignored which will be less
accurate for masonry in compression.
where is the shear stress, and n is the normal stress (compressive stress as a
negative quantity and tensile stress as a positive quantity), c and are the cohesion
and the angle of internal friction, respectively. The coulombs failure criterion is an
irregular hexagonal pyramid in the principal stress space. The cross-sectional shape
of this pyramid on the -plane is shown in Figure 2.30.
The Drucker-Prager criterion can be made to match with the apex of the Coulomb
criterion for either point A or B on its -plane as shown in Figure 2.30. For Point A,
where the cone circumscribes the hexagonal pyramid (the outer cone), the two
surfaces are made to coincide along the compressive meridian (Lode angle = /3),
and the Drucker-Prager parameters and k are related to the Coulomb constant c
and by
2 sin 6c cos
= ,k = (3)
3 (3 sin ) 3 (3 sin )
While for point B (the inner cone) the two surfaces are matched along the tensile
meridian (Lode angle = 0), and will have the constants
Sustainable Bridges SB-4.7.1 2007-11-30 58 (72)
2 sin 6c cos
= ,k = (4)
3 (3 + sin ) 3 (3 + sin )
As the material constants in ANSYS DP model are chosen to match with the
compressive meridian of the Coulomb criterion, therefore, the outer cone yield
surface is selected. The corresponding yield surface in p-q plane is shown in Figure
2.32, where p is the mean stress or hydrostatic stress and q is Mises equivalent
deviatoric stress.
q
Drucker-Prager yield criterion
6c cos
3-sin
c cos p
sin
The flow theory of plasticity is based on three basic assumptions: (1) the existence of
an initial yield surface; (2) the evolution of subsequent loading surfaces (hardening
rule); (3) the determination of an appropriate flow rule. For the current study, the
model is developed using Drucker-Prager material model implement in ANSYS,
therefore, the yield surface does not change with progressive yielding, hence there is
no hardening rule and the material is assumed elastic - perfectly plastic.
The material constants used is shown in Table 1. The effect of the weight of the soil
is neglected.
Table 2-2. Material properties for simple soil-arch model
Material constants
Youngs modulus N/mm2 207
Poissons Ratio 0.3
Cohesion N/mm2 0.069
Angle of internal friction degree 20
Both the associated and non-associated flow rule were adopted for the analyses of
the footing. The load-displacement relationship from the analyses is shown in Figure
2.34. The displacement vector using both associated flow rule and non-associated
flow rule are shown in Figure 2.35.
The collapse load using associated flow rule is around 2.5 MPa, more than twice that
of the load (152 psi or 1.05 MPa) predicted by the Coulomb criterion adopted by
Zienkiewicz et al., The load predicted by the Coulomb criterion is close to the loads
given by the Terzaghi and Prandtl solutions (1.2 MPa and 1.0 MPa).
Sustainable Bridges SB-4.7.1 2007-11-30 60 (72)
3
Applied pressure beneath footing (N/mm2)
2
non-associated flow rule
0
0 1 2 3 4 5
Displacem ent at center of footing (m m )
As a result, the analysis with the Drucker-Prager material constants matched with the
compressive meridian of the Coulomb criterion does not agree with the well-known
Terzaghi and Prandtl solutions. Studies by others (Griffiths, 1990) also suggested
that Drucker-Prager outer cone over-predicts the strength of soil. Therefore, the
selection of the material constants using the Drucker-Prager criterion is critical.
The arch and soil were both modelled using SOLID45 (8-node solid element), while
contact element CONTACT173 and target element TARGET170 (4-node interface
elements) are used for modelling the interface between soil and arch barrel interface.
A typical meshed simple soil-arch interaction model is shown in Figure 2.37.
h_soil
5
21
750
Figure 2.37: FE mesh, loading and boundary condition for simple soil-arch interaction
model
The boundary condition for the simple soil-arch interaction model is shown in Figure
2.37. The arch barrel was fixed at both supports, and the plain strain condition is
assumed for the soil.
A vertical patch load (width of 219mm) was applied at the quarter span position and
the maximum applied pressure 0.3N/mm2 is determined by controlling the maximum
tensile stress in the arch barrel which is no greater than the assumed tensile strength
of the material (0.5N/mm2 based on previous experimental data for similar brickwork
arches (Melbourne et al. 2007).
The material properties for the arch and the soil are given in Table 2-3. The
coefficient of friction for the interface is assumed to be 0.7 except for the study of the
influence of this parameter (varies from 0 to 1.0).
Table 2-3. Material properties for simple soil-arch model
Material Arch barrel Soil
Youngs modulus (N/mm2) 16000 1000*
Poissons Ratio 0.2 0.2
Density (Kg/ m3) 2200 2000
2
*varies from 200 to 10000N/mm for relative stiffness studies
In order to study the fixity conditions of abutments, the full bridge model had two
different abutment supports: one with abutments comprised top and bottom parts, the
other with the same size but only comprised one part.
500
300
215
2165
1665
220
750
240
530
730 322 3000 325 3923
8300
The arch was model using SOLID65 element with cracking and crushing capabilities,
while the soil was model using the same element but with Drucker-Prager properties.
The same interface elements as the simple soil-arch model are used for the full
bridge models. Due to the three-dimensional nature of the element SOLID65, three-
dimensional models were created. For simplicity and computing efficiency, only one
layer of elements were generated along the width of the model considering the load
was applied uniformly along the width of the bridge. A typical meshed full bridge
model is shown in Figure 2.39.
The boundary condition for the full bridge model is similar to that of simple soil-arch
interaction model except the soil was constrained by rigid tank which is fixed in all
direction and the lower section of the abutments are fixed at the bottom face. A
vertical patch load (width of 219mm) of the same size and same location as the full-
scale bridge test was applied incrementally
Table 2-4. Material properties for full bridge model
Arch barrel soil
2
Youngs modulus N/mm 16000 1000
Poissons Ratio 0.2 0.2
Density Kg/ m3 2200 1910
Uniaxial tensile cracking stress N/mm2 0.48
Uniaxial compressive stress N/ mm2 24
Cohesion N/ mm2 0.0224
Internal angle of friction degree 46.4
Sustainable Bridges SB-4.7.1 2007-11-30 63 (72)
S1max
Stress (N/mm2)
Stress (N/mm2)
0.5
S1max
0.4 SEQV
0.3
0.2
0.1
0
0 20 40 60 80
E_arch/E_soil
Figure 2.41: Influence of the relative stiffness between arch barrel and soil
Sustainable Bridges SB-4.7.1 2007-11-30 64 (72)
The influence of relative stiffness is clearly very significant not only on the value of
the maximum stresses but also on the positions of where these stresses occur. For
most case (except for the E_ratio < 4), the maximum tensile stress occurs in the
intrados of the arch barrel indicating that if the same yield criteria is used for both soil
and arch, the yielding or cracking will initiated from the arch barrel intrados.
The influence of contact stiffness on soil arch interface
Special care should be taken when selecting the values of contact stiffness between
arch barrel and soil interface. Normal contact stiffness is used to enforce compatibility
between the contact surfaces. If the contact stiffness factor (KFN) is too small, the
amount of penetration of contact surface into target surface may be too great resulted
in fictitious soft interface and the solution can be incorrect. On the other hand, if the
stiffness is too big, the determination of the true contact status normally requires
more iterations, and in some cases, convergence difficulties are inevitable.
The study has shown that the normal contact stiffness factor of 0.1 gives reasonable
stress levels in the arch barrel and prevent overlapping between interfaces
(penetration smaller than 0.1mm), it also leads to efficient solutions in terms of the
number of iterations.
The influence of mesh density
The art of using the finite element method lies in an appropriate mesh density to
solve a problem. If the mesh is too coarse then the inherent element approximations
will not allow a correct solution to be obtained. Alternatively, if the mesh is too fine the
cost of the analysis can be out of proportion to the results obtained. It is therefore
important to use a sufficiently refined mesh to ensure that the results from FE
simulation are adequate.
As far as the deflection is concerned even a coarse mesh will result in easy
convergence. For the convergence of stress, at least four elements are need across
the thickness of the arch barrel. For the maximum principal stress at an integration
point, the convergence is much slower and the difference between stress from
integration point and stress extrapolating from integration point always exists even
with a finer mesh (more than eight elements across the thickness). Mesh sensitivity
will, therefore, exist when cracking is based on the stress at integration points. The
rest of the studies are based on four elements across the thickness of the arch barrel.
The influence of the coefficient of friction on the interface
The influence of the coefficient of friction (MU) between the arch barrel and soil
interface is studied by varying the MU from 0 (frictionless contact) to 1.0 (rough
contact). The influence on stress in arch barrel and soil is shown in Figure 2.42. The
influence on the stress level is not very significant considering the contact interface
has changed from frictionless contact to relatively rough contact.
Sustainable Bridges SB-4.7.1 2007-11-30 65 (72)
0.6
0.5
Stress (N/mm2)
0.4
S1max
0.3
SEQV
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2
MU
Although the maximum horizontal movement on both side is less than 0.2mm at the
current load level, the maximum principal stress in the arch barrel has nearly tripled,
and the position of the stress has moved from quarter span intrados to the extrados
of left-hand abutment, which indicates that the crack will start from here rather than
the quarter span intrados.
the load causing the first cracks to occur would be lower. The load for the occurrence
of the first crack was 54kN/m for the abutment comprised two parts (top and bottom)
and 59kN/m for the corresponding bridge with abutment comprised one part.
Figure 2.45 show the displacement vector from the FE full bridge models with two
different abutment supports. Clearly with moveable abutment (abutment comprised
top and bottom parts), more soil was mobilised.
Figure 2.43: Principal stress (S1) distribution in the arch barrel and the bridge
(abutment comprised one part) (P=59kN/m)
Figure 2.44: Principal stress (S1) distribution in the arch barrel and the bridge
(abutment comprised two parts) (P=59kN/m)
Sustainable Bridges SB-4.7.1 2007-11-30 67 (72)
The bridges eventually failed by the formation of hinged mechanism plus the sliding
between top and bottom parts of the abutments. All these essential features are
similar to the large-scale bridge test.
The study has shown that the maximum load predict from the FE model is very
sensitive to the material properties such as cohesion, angle of friction, dilatancy,
tensile strength of the brickwork, density and relative stiffness between brickwork and
soil. Therefore, interpretation of these parameters from lab tests is critical.
Figure 2.46 shows an example of the influence of relative stiffness of the arch barrel
to the backfill. With relatively stiffer backfill, the bridge can carry much higher load
than that with softer backfill.
Figure 2.46: Cracks in the arch barrel for different stiffness of backfill (P=164kN/m)
Sustainable Bridges SB-4.7.1 2007-11-30 68 (72)
2.5 Conclusions
From the experimental programme, the following conclusions have been drawn:
Tests on pilot crushed limestone and clay filled bridges have been performed.
All the bridges tested failed by hinged mechanisms, although some abutment
movement was recorded in the tests;
The limestone filled arch bridges proved capable of carrying more load than
clay filled bridge. This demonstrates the importance of the fill type and
indicates that performing intrusive investigation to ascertain fill type is likely to
be worthwhile in the case of bridges with borderline load carrying capacity;
The tests indicated that the test rig performed as designed, with minimal frame
deflection (thereby giving effectively plane strain conditions);
Inclusion of large windows along one side of the test chamber permitted the
acquisition of good quality particle image velocimetry (PIV) data which is being
used to help better understand the nature of the soil-arch interaction.
The following conclusions can be drawn from the current finite element analyses:
The relative stiffness of the arch and the soil is a very important factor on soil-
arch interaction;
The fixity of the abutments affect the load at which the 1st crack occurs and
will in turn affect the load redistribution in the bridge;
The full bridge model can predict the essential feature observed in full scale
experiments;
The load carrying capacity from the FE model is very sensitive to the material
properties like density, cohesion of soil, the internal angle of friction, dilatancy
and the tensile strength of the brickwork. Therefore, interpretation of these
parameters from lab tests is critical.
Sustainable Bridges SB-4.7.1 2007-11-30 69 (72)
3. Acknowledgement
The authors wish to acknowledge the financial and technical support provided by
Network Rail and Essex County Council throughout the project.
The soil-arch interaction project is in collaboration with Sheffield University and the
authors acknowledge the financial support provided by Essex County Council and
Network Rail.
Sustainable Bridges SB-4.7.1 2007-11-30 70 (72)
4. References
Bazant, Z., Planas, J. (1997). Fracture and Size Effects in Concrete and Other
Quasi-brittle Materials, CRC Press, USA.
Boothby, T. E. and B. J. Roberts (2001). "Transverse behaviour of masonry arch
bridges." The Structure Engineer Volume 79(9): 21-26.
Burroughs, P., T. Hughes, S. Hee, and M. Davies (2002). "Passive pressure
development in masonry arch bridges." Proc. ICE, Structures & Buildings 152(4):
331-339.
Chen, W. F. (1985). Constitutive Relations for Concrete, Rock and Soils: Discusser's
Report. Mechanics of Geomaterials. Z. Bazant, John Wiley & Sons Ltd.: 65-73.
Chen, W. F. a. E. M. (1990). Non-linear Analysis in Soil Mechanics - Theory and
Implementation. The Netherlands, Elsevier Science Publishers B.V.
Edgell, G. (2005). Testing of Ceramics in Construction. Whittles Publishing,
Caithness, ISBN 1-870325-43-5.
Fairfield, C. a. D. P. (1994). "Model tests to determine the effect of fill on buried
arches." Proc. ICE Structures and Buildings 104(4): 471-482.
Fang, Y., Chen, T., Holtz, R. and Lee, W. (2004). "Reduction of boundary friction in
model tests." Geotechnical Testing Journal 27(1): 3-12.
Fanning, P. J., Boothby, T. E. (2001). Non-linear three dimensional simulations of
service load tests on a 32m stone arch bridge in Ireland. Arch'01 Third international
arch bridges conference, Paris.
Garrity, S. W. and Toropova, I. L. (2001). A finite element study of a single span
masonry arch bridge with near surface reinforcement. Arch'01 Third international
arch bridges conference, Paris.
Gilbert, M. Smith, C.C., Melbourne. C and Wang. J . (2006). An experimental study of
soil-arch interaction in masonry bridges. IABMAS'06 - Third International Conference
on Bridge Maintenance, Safety and Management, Portugal.
Griffiths, D. V. (1990). "Failure criterion interpretation based on Mohr-Coulomb
friction." J. Geotech. Engng. ASCE(116): 986-999.
Heyman, J. (1966). The stone skeleton. Structural Engineering of Masonry
Architecture. University of Cambridge, ISBN-13 9780521629638, ISBN-10
0521629632.
Heyman, J. (1996). Arches, Vaults and Buttresses, Masonry Structures and their
Engineering Valorium. ISBN 0860785971.
Highway structures: Design (Substructures and Special Structures), Materials.
Special Structures. Unreinforced Masonry Arch Bridges. DMRB Volume 2 Section 2
Part 14 (BD 91/04), Highways Agency, The Stationary Office, London.
Highway structures: Inspection and maintenance, Inspection, Advice Notes on the
Non-destructive Testing of Highway Structures. (2004). DMRB Volume 3 Section 1
Sustainable Bridges SB-4.7.1 2007-11-30 71 (72)
Part 7 (BA 86/04), Highways Agency, The Stationary Office, London, Highways
Agency Advice Note BA86.
Hogg, V. (1997). Effects of Repeated Loading on Masonry Arch Bridges and
Implications for Serviceability Limit State. PhD Thesis, University of Nottingham.
Hughes, T.G., Davies, M.C.R., Taunton, P.R. (1998). The Small Scale Modelling of
Masonry Arch Bridges Using a Centrifuge. Proc. Inst. Civ. Engrs., Structures &
Building Journal, 128, pp. 49-58.
Lange, D.A., Jennings, H.M., Shah, S.P. (1993). Relationship Between Fracture
Surface Roughness and Fracture Behavior of Cement Paste and Mortar. Journal of
American Ceramic Society, 79, (3), 589-597.
Lourenco, P. B. (1996). Computational Strategies for Masonry Structures, PhD
thesis, Delft University of Technology.
Melbourne, C. and Gilbert, M. (1995). "The behaviour of multi-ring brickwork arch
bridges." The Structural Engineer 73(3): 39-47.
Melbourne, C., Tomor, A.K., Wang, J., (2005). Modes of Failure of Multi-ring
Masonry Arches under Fatigue Loading. Proc. Bridge Management Five
Conference, UK, pp 476-483.
Melbourne, C., Tomor, A.K., (2006). A New Assessment Method for Masonry Arch
Bridges. Proc. IABMAS06 Conference, Porto, Portugal.
Melbourne, C. Wang, J. and Tomor, A. (2007). "New Masonry Arch Bridge
Assessment Strategy (SMART)." Proc. ICE, Bridge Engineering 160: 81-87.
Tomor, A.K., Melbourne, C. (2007). Monitoring Masonry Arch Bridge Response to
Traffic Loading Using Acoustic Emission Techniques. Proc. ARCH07 Conference,
Madeira.
Tomor, A.K., Melbourne, C., (2007). Condition Monitoring of Masonry Arch Bridges
Using Acoustic Emission Techniques, Structural Engineering International (IABSE),
2/2007.
Page, J. (1993). Masonry Arch Bridges. London, HMSO.
Page, J. (1995). (1995). Load Tests to Collapse on Masonry Arch Bridges. Thomas
Telford, London, pp. 289-298.
Pippard, A J.S. (1948). The Approximate Estimation of Safe Loads on Masonry Arch
Bridges. Civil Engineer in War, 1, 365-372, ICE, London.
Pippard, A.J.S, Chitty, L. (1951). A Study of the Voussior Arch. National Building
Series, Research Paper 11, London.
Tao, H. (2001). The behaviour of open spandrel brickwork masonry arch bridges.
PhD thesis, University of Salford.
Wang, J. (2004). Three dimensional Behaviour of Masonry Arches. PhD thesis,
University of Salford.
Wang. J., Smith, C.C., Gilbert, M. and Melbourne. C (2006). Load test to collapse of
back-filled brickwork masonry arch bridges. Data Report: Salford-Soil-Arch-0001 and
Salford-Soil-Arch-0002.
Sustainable Bridges SB-4.7.1 2007-11-30 72 (72)
White, D. a. T., W. (2002). GeoPIV particle image velocimetry (PIV) software for use
in geotechnical testing. Technical Report 322. Cambridge, Geotechnics Group,
Cambridge University Engineering Department.
Willam, K. J. and E. D. Warnke (1975). Constitutive Model for the Triaxial Behavior of
Concrete. Proceedings, International Association for Bridge and Structural
Engineering, ISMES. Bergamo, Italy, ISMES. Vol. 19: 174.
Zienkiewicz, O. C., Humpheson, C. and Lewis, R. W. (1975). "Associated and non-
associated visco-plasticity and plasticity in soil mechanics." Geotechnique 25(4): 671-
689.
Numerical Analyses of Load Distribution and
Deflections in Railway Bridge Transition Zones due to
Passing Trains
Background document D4.7.2
PRIORITY 6
SUSTAINABLE DEVELOPMENT
GLOBAL CHANGE & ECOSYSTEMS
INTEGRATED PROJECT
Sustainable Bridges SB-4.7.2 2007-11-30 2 (46)
This report is one of the deliverables from the Integrated Research Project Sustainable Bridges - Assessment for
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework Pro-
gramme. The Project aims to help European railways to meet increasing transportation demands, which can only
be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
passenger trains. This requires that the existing bridges within the network have to be upgraded without causing
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and econ-
omy of the railways.
A consortium, consisting of 32 partners drawn from railway bridge owners, consultants, contractors, research
institutes and universities, has carried out the Project, which has a gross budget of more than 10 million Euros.
The European Commission has provided substantial funding, with the balancing funding has been coming from
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule Tech-
nical University has undertaken the scientific leadership.
The Project has developed improved procedures and methods for inspection, testing, monitoring and condition
assessment, of railway bridges. Furthermore, it has developed advanced methodologies for assessing the safe
carrying capacity of bridges and better engineering solutions for repair and strengthening of bridges that are found
to be in need of attention.
The authors of this report have used their best endeavours to ensure that the information presented here is of the
highest quality. However, no liability can be accepted by the authors for any loss caused by its use.
PREFACE ................................................................................................................................1
SUMMARY...............................................................................................................................2
1 INTRODUCTION...............................................................................................................4
2 BACKGROUND................................................................................................................5
2.1 Railway Traffic on Soft Soil Sites ...............................................................................5
2.2 Railway Traffic in Bridge Transition Zones ................................................................6
3 NUMERICAL VERIFICATION STUDY .............................................................................9
3.1 Embankment and Instrumentation .............................................................................9
3.2 FLAC3D Model ........................................................................................................10
3.3 Results of FLAC3D Analyses...................................................................................12
3.4 Closing Remarks......................................................................................................14
4 NUMERICAL ANALYSES WITH CONCRETE BRIDGE................................................15
4.1 FLAC3D Model and Analyses..................................................................................15
4.2 Results of Parametric Analyses ...............................................................................17
4.2.1 Train Axle Load.................................................................................................17
4.2.2 Train Velocity ....................................................................................................21
4.2.3 Material Property Values of Ballast, Subballast, and Embankment Fill ............24
4.3 Summary and Conclusions ......................................................................................30
5 NUMERICAL ANALYSES WITH MASONRY ARCH BRIDGE ......................................31
5.1 FLAC3D Model and Analyses..................................................................................31
5.2 Results of Analyses .................................................................................................32
5.3 Summary and Conclusions ......................................................................................38
6 LIMITATIONS .................................................................................................................40
7 CONCLUSIONS..............................................................................................................41
8 REFERENCES................................................................................................................42
Sustainable Bridges SB-4.7.2 2007-11-30 1 (46)
PREFACE
For the assessment of existing railway bridges it is important to have a holistic approach that
takes into account the bridge, foundation and the embankment transition zones. The load
distribution and deflections in the transition zones due to passing trains, which will be faster
and heavier in the future, should be assessed with respect to the influence on the bridge.
The main development achieved in this work package of the Sustainable Bridges project are
numerical analyses used to evaluate the effect of train velocity and load, as well as the stiff-
ness of the ballast, subballast, and embankment material on the stresses and deflections in
the transition zones. Three dimensional dynamic analyses were performed to study two
bridges representative of existing European railway bridges, one concrete bridge and one
masonry arch bridge. No laboratory or field tests have been performed. However a numeri-
cal verification study was performed using the measured data at a full-scale instrumented
railway embankment in Finland (Kolisoja and Mrkel, 2001). This study shows good agree-
ment between the 3D numerical analyses and pressure cell data. The obtained results indi-
cate that 3D numerical analyses can be a useful tool when evaluating the load distribution
and deflections in transition zones due to passing trains. The findings of the numerical
analyses should be verified against laboratory scaled tests and/or full-scale field tests.
The technical development and this report have been financed by the Sustainable Bridge
project.
Sustainable Bridges SB-4.7.2 2007-11-30 2 (46)
SUMMARY
Many railway lines in Europe are 60 to 100 years old, and are not designed in accordance with
modern railway traffic. As identified by many railway authorities, faster and heavier modern
trains are causing problems at existing railway bridges. Bridge ends are one of the most com-
plicated parts of a railway track since, at these locations, the rail is subjected to concentrated
stresses, which are due to dynamic and cyclic loads caused by the passage of trains and
bridge deflections. The extent of the problems that develop within the transition zone is de-
pendent upon the train load and velocity, and subgrade stiffness. Unlike train traffic on soft
sites, little has been published about the effect of train traffic on the load distribution and de-
flections in the bridge transition zone.
The scope of this work included performing numerical analyses to evaluate the load distribu-
tion and deflections in bridge transition zones due to the passing of trains. No field or labora-
tory tests were performed. This document is a background document, intended to provide
insight into the general behavior of bridge systems under the passing of high speed trains.
No design methods or recommendations are presented herein.
The greatest increases in vertical and horizontal stresses are concentrated within the
upper 1.0 to 1.2 m of the system. Vertical and horizontal stresses caused by the pass-
ing of a train decrease with depth.
As train axle loads increase, the vertical and horizontal stresses, and vertical deflec-
tions beneath the train increase.
As train velocity increases, the horizontal stresses and vertical deflections increase,
however, they do not increase linearly with increasing velocity. As the velocity of the
train approaches the Rayleigh wave velocity of the embankment system, significant
horizontal stresses and vertical deflections develop.
As the stiffness of the ballast and/or subballast materials increase, net horizontal
stresses exerted on the bridge decrease. Horizontal stresses in the embankment fill
are relatively small, but increase when a soft fill is used.
When a soft embankment fill is used, vertical deflections increase substantially, and
these deflections are observed to occur 3 m behind the train, as if there is a stern
wave behind the train.
Numerical analyses were also performed using a bridge geometry representative of Euro-
pean masonry arch bridges. It can be assumed that the factors that affect the stress distribu-
Sustainable Bridges SB-4.7.2 2007-11-30 3 (46)
tion and deflections for the concrete bridge structure (as described above) will have the same
effect for the masonry arch bridge structure. However, the geometry of the arch bridge is
different than the concrete bridge in two significant ways: (1) for masonry arch bridges there
is a gradual change in stiffness, while for a concrete bridge the change is more abrupt, and
(2) for the masonry arch structure, the ballast and subballast are confined within a relatively
small volume. Due to the latter factor, additional vertical and horizontal stresses induced by
the train are distributed immediately to the arch structure. Since there is less soil above the
crown of the arch to dissipate stresses, the crown of the arch will experience the greatest
increase in vertical stress under the passing of a train.
Sustainable Bridges SB-4.7.2 2007-11-30 4 (46)
1 INTRODUCTION
Many railway lines in Europe are 60 to 100 years old, and are not designed in accordance
with modern railway traffic. Due to faster and heavier modern trains, existing railway bridges
are experiencing problems. These issues have an adverse effect on the safety, reliability,
and economy of the railway line, and therefore, many existing bridge systems require up-
grading. Engineers are faced with the task of assessing the performance of existing bridges,
and, if necessary, designing the strengthening or repair systems.
The scope of this work included performing numerical analyses to evaluate the load distribu-
tion and deflections in bridge transition zones due to the passing of trains. No field or labora-
tory tests were performed. Numerical analyses consisted of three-dimensional dynamic
analyses. The numerical models are representative of two common European bridge types,
a concrete and a masonry arch bridge.
Sustainable Bridges SB-4.7.2 2007-11-30 5 (46)
2 BACKGROUND
Kopf and Adam (2005) performed a parametric study using an analytical model they devel-
oped. Kopf and Adam (2005) observed that the critical velocity of the system decreases with
increasing train load. They also observed that in the static case, the maximum deflection
occurs beneath the applied load; whereas in the dynamic case, there is a time lag between
the maximum applied force and the maximum deflection. As the train velocity increases, the
wave length of the bow wave in front of the load becomes shorter while the rear wave be-
hind the load becomes longer (Kopf and Adam 2005).
Kopf and Adam (2005) also present the results of field tests on a high speed railway line be-
tween Vienna and Salzburg. Pressure cells were placed beneath the railway sleepers, and a
train was passed over the tracks at a range of velocities up to 230 km/hr. The pressure cell
measurements revealed that the maximum dynamic pressure beneath the train increased
with increasing train velocity.
Katzenbach and Ittershagen (2005) describe field tests that were performed to evaluate the
dynamic soil-structure interaction of railway lines on soft soil sites. Stress measurements were
taken within the soil at depths of 0.4 m beneath the sleepers under the passing of passenger
(axle load = 100 kN) and freight (axle load = 225 kN) trains. Katzenbach and Ittershagen
(2005) observed that each crossing of a train axle caused a single stress amplitude. Stress
amplitudes were less for the passenger train, which had smaller axle loads, than the freight
train. Measured particle velocities decreased with depth, and increased with increasing train
velocity (Katzenbach and Ittershagen 2005).
Adolfsson et al. (1999a) and Adolfsson et al. (1999b) describe a project funded by Banverket
to evaluate deflections and accelerations in soft soil due to the passing of trains with veloci-
ties up to 204 km/hr. The project site is located in Ledsgrd, approximately 25 km south of
Gothenburg, Sweden. An X2000 train set was used and the train velocity was varied from
zero (static) to 204 km/hr. Axle loads were on the order of 120 to 190 kN. The embankment
at the project site consisted of 0.5 m of ballast underlain by 0.9 m of gravel. Below the em-
bankment, the soil conditions consisted of 3 m of gyttja (organic clay) underlain by more than
50 m of soft clay. The shear wave velocity of the soft soil increased from 45 m/s to 90 m/s at
a depth of 14 m (Adolfsson et al. 1999a). The railway and embankment were instrumented
with displacement transducers and accelerometers. There was no bridge at the project site.
Sustainable Bridges SB-4.7.2 2007-11-30 6 (46)
Based on the measurements at the Ledsgrd site, Adolfsson et al. (1999a) observed that, at
low train velocities, the vertical deflections were static and occurred in the upper 5 m of the
soil. At train velocities 200 km/hr or greater, the deflections were approximately three times
larger and occurred to depths up to 10 m. Adolfsson et al. (1999b) also observed that the
maximum downward deflections usually occurred below the bogie axles of the heaviest
wagon of the train set, which is the engine wagon for the X2000 train. Adolfsson et al.
(1999b) point out that, for higher train velocities, a traveling wave in the embankment is de-
tected by the instrumentation behind the train, much like a stern wave behind the ship.
Numerical analyses described by Adolfsson et al. (1999b) confirm the tail wave phenomena.
The numerical analyses show the shifting from a static-like deflection pattern of the ground
surface at low train velocities to a plough-shaped dynamic pattern at high train velocities
(Adolfsson et al. 1999b).
Furthermore, the high vibration levels detected at the Ledsgrd site were detected under the
passing of the X2000 train but not under the passing of cargo trains. Therefore, it was de-
duced that the vibration problems were related to the high velocity of the X2000 train, and not
the train axle loads (Adolfsson et al. 1999a).
Lundqvist and Dahlberg (2004) note that track settlements are often associated with the dy-
namic loading of high-speed train traffic, and that high-speed railway traffic has caused track
settlements, even if the high-speed trains have relatively low axle loads. However, based an
analytical model presented by Hunt (1997), the total magnitude of the track settlement is
governed largely by the magnitude of the axle loads, although some local effects are con-
trolled by dynamic loads. Hunt (1997) noted that for train velocities up to 350 km/hr, the ve-
locity of the wave propagation in most tracks is very much greater than the train velocity.
The response of the track system is highly dependent upon the Rayleigh-wave velocity of the
soil and structural components. As a rule of thumb, Madshus et al. (2004) recommend that
track systems have a critical speed of at least 1.7 times the operating maximum speed of the
train. They note that there are basically two ways to increase the critical speed of the track
structure-ground response: (1) by improving the stiffness and strength of the ground be-
neath the track structure, and (2) by increasing the longitudinal bending stiffness and thus
increasing the bending wave speed along the track itself.
Many of the problems that occur within the transition zones are associated with modern train
traffic, which is heavier and faster than the traffic for which the bridges were designed. Due to
faster and heavier modern trains, existing railway bridges are experiencing problems, such as
an increase of differential settlements within the bridge transition zone, and an increase of
loads on existing bridge structures. An illustration of problems that can develop in bridge tran-
sition zones is shown schematically in Figure 2.1.
Sustainable Bridges SB-4.7.2 2007-11-30 7 (46)
Train
Ballast Layer
Bridge Structure Increased vertical deformations
Increased vertical and
horizontal stresses behind
the bridge
Embankment
Subsoil
ERRI (1999) developed a State of the Art report based on the responses of 14 participating
European railways to a questionnaire relating railway track, bridges, and earthwork. A few of
the important conclusions made by ERRI (1999) include:
There is insufficient coordination between track, bridge and geotechnical engineers.
In many countries, the transition zones of existing lines give rise to more problems
and require more maintenance than do new lines.
There is an urgent need to develop methods for upgrading existing transitions zones
which can be implemented with minimum disturbance to traffic.
ERRI (1999) states that maintenance within the transition zone is required five times more
than for other track locations. It is unknown how much of this required maintenance is al-
ready attributed to heavier and faster modern railway traffic; however, it can be expected that
as train loads and velocities increase, the amount of required maintenance will also increase.
Settlements have been identified as the major contributory factor in deterioration of the track
geometry in bridge transition zones (ERRI 1999). Settlements may occur in the subsoil and/or
in the bridge backfill material. Li and Davis (2005) identified three major causes of settlement
within the transition zones:
1. A large difference in stiffness between the portion of the track on the bridge and the por-
tion in the transition zone can lead to uneven deflections;
2. Unlike the bridge, which is supported on stiff foundations, the transition zone is sup-
ported on the subsoil and inherently settles more than the bridge;
3. Settlements develop due to poor embankment materials, inadequate compaction, and
poor drainage conditions.
These problems have an adverse effect on the safety, reliability, and economy of the railway
line. Differential settlements that develop within the transition zone have several impacts on
the railway operations, such as required repeated maintenance work and restrictions on train
velocities. As soon as differential settlements begin to develop, the variations of the dynamic
train/track forces increase and this speeds up the track deterioration process (Lundqvist and
Dahlberg 2004).
Sustainable Bridges SB-4.7.2 2007-11-30 8 (46)
Li and Wong (2005) investigated four bridge sites in Kansas, USA. The bridges are precast
concrete ballasted deck bridges and are 9 to 54 m long, with 8.4 m spans typical. All four
bridges are located on the same track route that sees a significant amount of traffic, with
70% of the traffic having axle loads of 36 tons. A discussion of the soil conditions is provided
by the authors. Li and Wong (2005) observed that the tracks in the transition zone experi-
enced more degradation than the tracks on the bridge or the tracks far away from the bridge.
Li and Wong (2005) also concluded that, for the four sites investigated, the resulting differen-
tial settlement within the transition zone came mostly from the ballast and subballast layers,
with some additional contribution from the underlying soil layers. The authors note that reme-
dies intended to strengthen the subgrade were not as effective because they did not com-
pletely address the stiffness issues associated with the bridges and the transition zones. Li
and Wong (2005) recommend that mitigation techniques such as rubber pads under the con-
crete sleepers or rubber mats on the concrete bridges be used to reduce track stiffness on
bridges, to increase damping characteristics, and to reduce impacts between the ballast parti-
cles and concrete sleepers or bridge surfaces.
Olofsson and Hakami (2000) performed three-dimensional numerical analyses to evaluate the
interaction between the train track, bridge abutment, and backfill in the transition zone. Based
on the analyses performed by Olofsson and Hakami (2000), the stiffness of the embankment
fill had a significant effect on the deflections within the transition zone. As the stiffness of the
backfill decreased, differential deflections increased. However, calculated vertical deflections
on the bridge platform were negligible.
As identified by many railway authorities (ERRI 1999, Li and Wong 2005), faster and heavier
modern trains are causing problems at existing railway bridges. However, unlike train traffic
on soft sites, less has been published about the effect of train traffic on the load distribution
and deflections in the bridge transition zone.
Sustainable Bridges SB-4.7.2 2007-11-30 9 (46)
The field tests were analyzed to verify the use of FLAC3D models to evaluate the load transfer
behaviour of embankments subjected to moving train loads. Results from the FLAC3D analy-
ses were compared to (1) the published results of the field measurements, and (2) the results
of two-dimensional linear elastic analyses performed by Kolisoja et al. (2000) of their own field
tests. A discussion of the behaviour observed by Kolisoja and Mrkel (2001) during the field
tests is also provided.
Figure 3.1. Schematic cross section of embankment (from Kolisoja and Mkel 2001)
Instrumentation consisted of 16 strain transducers and four pressure cells. The pressure cells
were installed at depths of 0.5 and 1.0 m below the bottom of the sleepers, and the strain
transducers were installed at depths of 0.5, 1.0 and 1.8 m. Measurements were taken over a
period of 44 hours during July 1999. Normal train traffic was maintained, and in addition, a
test train with four axle loads of exactly 250 kN passed over the instrumented site ten times at
velocities of 40 to 100 km/hr.
Sustainable Bridges SB-4.7.2 2007-11-30 10 (46)
8.2 m 3.3 m
C
L 2.5 m
Ballast 1.0 m
Sand 0.8 m 0.4 m
Gravel
Clay
4.65 m
30.0 m
The model is 30.0 m wide and 25.0 m long. Zone sizes are on average 0.25 x 0.27 x 0.26 m,
and a total of 129,200 zones were used. Previous mesh refinement studies have shown that
this mesh refinement was sufficient to reach numerical convergence. In FLAC3D, the spatial
element size should be smaller than approximately one-tenth to one-eighth of the wavelength
associated with the highest frequency component of the input wave (Itasca 2005). For the
analyses presented herein, the wavelength of the input wave is equal to 7.25 m, which corre-
sponds to the length of the simplified load associated with one bogie.
The approximated load representing the load beneath two wheels of one bogie is schemati-
cally shown in Figure 3.3. Numerical problems can develop when attempting to apply the
moving train load as successive point loads on structural elements (i.e. rails and sleepers);
therefore, the train load was input as a cosine-wave. The cosine-wave was applied succes-
sively over each zone of the model. The duration of the cosine-wave is equal to the length of
one bogie (7.25 m) divided by the velocity of the train. The width and magnitude of the wave
are determined based upon the theory of a loaded beam on a Winkler elastic bed, which
represents the distribution of the train load to the system by the sleepers. Granted, this load
representation is a simplification but it has been found to reasonably represent the actual
train load in analysing the high-speed phenomenon by other researchers (e.g. Andrasson
2002). This section describes further verification of the FLAC3D modelling procedure
against field data.
Sustainable Bridges SB-4.7.2 2007-11-30 11 (46)
2.9 m
The initial train load was calculated based upon an axle load of 250 kN. Axle loads of 125,
225, and 330 kN were also considered in the study. For all of the FLAC3D analyses in this
verification study, a train velocity of 50 km/hr was used.
The material property values used in the numerical analyses are listed in Table 3.1. A linear-
elastic perfectly plastic model with a Mohr-Coulomb failure criterion was used to represent
the ballast, sand, gravel, and clay materials. The elastic modulus of the ballast material de-
creases with depth. In the FLAC3D model, the ballast layer was divided into four sublayers
with values of elastic modulus equal to 400, 250, 175, and 155 (MPa), respectively. These
values of elastic modulus are equal to the values used by Kolisoja et al. (2000) in their two-
dimensional analyses.
Table 3.1. Material property values used in the analyses (from Kolisoja et al. 2000)
Moist Friction
Poisson's Elastic Modulus Cohesion
Material Density Angle
Ratio (MPa) (kPa)
(kN/m3) (deg)
Decreases from 400 to
Ballast 22 0.3 40 0.1
155 with depth
Sand 18 0.3 145 33 0.1
Gravel 20 0.3 160 35 0.1
Clay 16 0.3 40 20 10
No material damping was used. Additionally, no structural elements were used to model the
rail and sleepers. It was desired to model the train dynamically, and since the timestep in the
dynamic mode is determined by the largest material stiffness, using structural elements
would require unrealistically long model run times, and may cause numerical problems.
Sustainable Bridges SB-4.7.2 2007-11-30 12 (46)
The vertical pressure with depth beneath a train with an axle load of 250 kN is shown in Figure
3.4. Instrumentation data was provided at discrete points beneath the train load; data of the
longitudinal distribution at each depth is not available. As can be seen in Figure 3.4, both the
field measurements and numerical analyses indicate that the additional vertical pressures
caused by the passing of a train decrease with depth. The greatest increase in vertical stress
is concentrated within the upper meter of the embankment. Kolisoja and Mkel (2001) point
out that one pressure cell appears to give an unreliable measurement; however, in general,
the authors state that the instrumentation performed well and that meaningful results were
obtained even from layers that, at the beginning, were considered to be too coarse grained for
any instrumentation.
-0.5
Depth from the Bottom of Sleeper (m)
-1
-1.5
-2.5
Figure 3.4. Vertical pressure with depth for a train axle load of 250 kN
The pressure cell measurements at a depth of 0.5 m below the bottom of the sleepers for vari-
ous axle loads are shown in Figure 3.5. The data in Figure 3.5 include measurements taken
beneath four crossings of the test train with an axle load of 250 kN, various tank trains, and
one crossing of maintenance equipment with four exceptionally light axle loads (Kolisoja and
Mkel 2001). As can be seen in Figure 3.5, there is a relatively linear trend between the
measured vertical pressure and axle load, though there is some scatter in the data. The re-
sults of the two-dimensional numerical analyses by Kolisoja et al. (2000) are shown on Figure
3.5 as the black circles. Data was also presented in Finnish by Kolisoja et al. (2000) for pres-
sure cell measurements taken at a depth of 1.0 m below the bottom of the sleepers.
Sustainable Bridges SB-4.7.2 2007-11-30 13 (46)
100
Vertical Pressure (kPa)
80
20
0
50 100 150 200 250 300 350
Axle Load (kN)
Figure 3.6. Vertical pressure at depths of 0.5 and 1.0 m as a function of axle load
The vertical pressures calculated by FLAC3D at depths of 0.5 and 1.0 m below the bottom of
the sleeper for various axle loads are presented in Figure 3.6. The solid lines in Figure 3.6
are the results of the analyses performed by Kolisoja et al. (2000), which also seem to serve
as "best fit" lines for their field measurements (see Figure 3.5). As can be seen in Figure 3.6,
there exists a fairly linear relationship between vertical pressure and train axle load for the
Sustainable Bridges SB-4.7.2 2007-11-30 14 (46)
range of axle loads investigated. Also, it can be seen here that vertical pressures decrease
with depth. Furthermore, as the train load increases, the difference between the stresses at
a depth of 0.5 m and 1.0 m increase. It is reasonable to assume that for the range of axle
loads investigated, the greatest increase in vertical stresses remain concentrated in the up-
per meter of embankment material.
As can be seen in Figures 3.4 and 3.6, the FLAC3D analyses provide a good fit to both the
field pressure cell measurements and the analyses performed by Kolisoja et al. (2000).
(1) the vertical stresses beneath a train decrease with depth, with most of the increase in
vertical stress concentrated within a depth of 1.0 m below the bottom of the sleeper,
and
(2) vertical pressures within the embankment system increase with increasing train axle
load.
The FLAC3D analyses provided a good fit to the pressure cell data. This verification study
demonstrated that:
(1) Reasonable results are achieved when a single bogie is considered, and the train
load is input dynamically as a cosine wave, and
(2) Though in reality the sleepers distribute the train load to the embankment system,
reasonable results were achieved in the numerical analyses despite the fact that no
structural elements were used to represent the rail or sleepers.
At low train velocities, the ground response from a moving load is essentially quasi-static;
however, stresses and vertical movements increase substantially as the train velocity ap-
proaches the Rayleigh-wave velocity of the soil (Adolfsson et al. 1999b, Kaynia et al. 2000,
Chen et al. 2005). The field measurements described by Kolisoja and Mkel (2001) in-
volved the use of trains with velocities ranging from 40 to 100 km/hr. At these velocities, the
Rayleigh-wave velocity of the embankment system was not surpassed; therefore, the system
responded in a quasi-static manner. As shown in the following section, as train velocities
approach the Rayleigh-wave velocity of the embankment system, deformations and stresses
increase non-linearly.
Sustainable Bridges SB-4.7.2 2007-11-30 15 (46)
The FLAC3D analyses described in the previous section provided verification and an under-
standing of the appropriate numerical modelling procedures for a railway embankment system
subjected to moving train loads. The same numerical modelling procedures were adopted for
the case in which a bridge exists along the railway. The aim of the analyses is to evaluate the
stress distribution and deflections within the transition zone of a railway bridge due to the
passing of train with high speeds and high axle loads.
A numerical parametric study was performed to study the effects of 1.) the train load, 2.) the
train velocity, and 3.) the stiffness of the ballast, subballast and embankment fill. The ap-
proach adopted for the numerical parameter studies relies on a base case analysis, with sys-
tematic variation of parameter values from the base case. The parametric study was per-
formed using a typical concrete bridge geometry. This section provides a description of the
results of the numerical parameter study, including a discussion of the trends disclosed by the
analyses.
The embankment profile of the bridge geometry is shown in Figure 4.2. The profile consists
of an upper ballast layer 0.4 m thick, which is underlain by a layer of subballast 1.2 m thick,
which in turn is underlain by embankment backfill. A stiff, non-yielding, bearing layer is as-
sumed to exist beneath the embankment fill. As can be seen in Figure 4.2, the profile is
symmetrical about its centre-line.
7.4 m
3.3 m Train Load
2.5 m
Ballast
0.4 m
Subballast 2 1.2 m
1
Embankment Fill
3.8 m
The material property values used in the numerical analyses are listed in Table 4.1. A linear-
elastic perfectly plastic model with a Mohr-Coulomb failure criterion was used to represent the
ballast, subballast, and embankment fill.
No structural elements were used to model the bridge abutments, or the rail and sleepers. It
was desired to model the train dynamically, and since the timestep in the dynamic mode is
determined by the largest material stiffness, using structural elements would require unrealis-
tically long model run times, and may cause numerical problems. Therefore, the bridge struc-
ture was represented by fixed gridpoints at the bridge-soil interface. Since deflections at the
structure beneath the passing train are on the order of a few millimetres, these boundary
conditions are not expected to significantly affect the results.
Sustainable Bridges SB-4.7.2 2007-11-30 17 (46)
The train load was applied dynamically along the top of the model over a width of 1.25 m from
the centre-line. To gain a clear understanding of the behaviour of the system under a dynamic
load, only one bogie of the train was considered. The simplified approximation of the load be-
neath two wheels of one bogie is similar to that shown in Figure 3.3, depending upon axle
spacing.
Results are presented at the first bridge abutment when the train is directly above the back of
the abutment (see Figure 4.3). The first abutment is the end of the bridge first encountered
by the train as the train moves toward and over the bridge. Results are presented for the first
abutment because there are significant increases in stresses exerted on the bridge, and de-
flections behind the bridge, that occur as the train approaches the bridge. This is due to the
fact that the train is moving from the railway embankment onto the stiff bridge structure. The
bridge structure is less affected by the moving train as the train moves from the bridge struc-
ture onto the railway embankment.
Position of bogie
for which results
are reported
Typically, the axle loads of freight trains are greater and the wheel spacings are less than
those of passenger trains. The passenger train loads were applied over a range of velocities
from 50 to 350 km/hr, and the freight trains were applied at velocities of 50 and 100 km/hr. A
velocity of 100 km/hr is considered to be an upper limit velocity for a freight train.
Sustainable Bridges SB-4.7.2 2007-11-30 18 (46)
The increase in horizontal stresses in the embankment fill material occurs slightly ahead of
the moving train. However, the greatest increases in horizontal and vertical stresses occur in
the ballast and subballast layers; these increases occur approximately beneath the train
when the train is above the back of the abutment. A cross-section of the net horizontal
Sustainable Bridges SB-4.7.2 2007-11-30 19 (46)
stresses on the first bridge abutment for axle loads of 125 and 250 kN are shown in Figure
4.4 for the case in the train velocity was equal to 350 km/hr.
As can be seen in Figure 4.4, the maximum net horizontal stresses occur between the depths
of 0.2 to 0.45 m. For all the analyses in this parametric study, the net maximum horizontal
stresses were observed within this range of depth, which means that in all cases, the maxi-
mum net horizontal stresses occur in the ballast layer or at the ballast-subballast layer inter-
face. For both cases, below a depth of about 1 to 1.2 m, the net horizontal stresses in the
embankment fill are relatively small. However, since the increase in horizontal stresses in the
embankment fill material occurs slightly ahead of the moving train, horizontal stresses at this
depth are slightly greater as the train is approaching the abutment.
The FLAC3D analyses indicate that as the train axle load increases, the vertical and horizontal
stresses within the railway system increase. Figure 4.5 shows the vertical pressures behind
the first concrete bridge abutment calculated by FLAC3D at depths of 0.2 and 0.4 m for vari-
ous axle loads traveling at a velocity of 50 km/hr. Figure 4.6 shows the horizontal pressures
calculated at depths of 0.2 and 0.4 m for (1) various axle loads traveling at a velocity of 50
km/hr and (2) axle loads of 125 and 250 kN traveling at 350 km/hr. It should be noted that the
ballast layer is 0.4 m thick, and the top of the bridge structure is located at the base of the bal-
last layer.
As can be seen in Figure 4.5, the vertical pressures are greater at the shallower depth of 0.2
m, which is expected. In Figure 4.6, it can be seen that there is not much difference in the
horizontal pressures between the depths of 0.2 and 0.4 m for a given train velocity; however,
horizontal pressures increase as train velocity increases. The effects of train velocity on verti-
cal and horizontal stresses are discussed in greater detail in the following section.
100 z = -0.2 m
z = -0.4 m
Vertical Pressure (kPa)
80
60
40
20
0
100 150 200 250 300 350
Axle Load (kN)
80
40
v = 50 km/hr, z = -0.2 m
20
v = 50 km/hr, z = -0.4 m
v = 350 km/hr, z = -0.2 m
v = 350 km/hr, z = -0.4 m
0
100 150 200 250 300 350
Axle Load (kN)
A cross-section of the vertical deflections 1 m behind the first bridge abutment is shown in Fig-
ure 4.7 for the case in which the train axle load was equal to 250 kN, the train velocity was
equal to 350 km/hr, and the base case material properties were used. When the train is di-
rectly above the back of the abutment, the maximum vertical deflections occur approximately
0.6 to 1 m behind the train (unless noted otherwise). For the case shown in Figure 4.7, the
maximum vertical deflection behind the first bridge abutment is 2.87 mm.
A few general observations can be made based on the results shown in Figure 4.7. First,
most of the vertical deflections occur in the ballast and subballast layers. Second, as would be
expected, the maximum vertical deflections are concentrated beneath the width of the train.
As the train axle load increases, the maximum vertical deflections increase. The vertical de-
flections for axle loads of 125 and 250 kN for a wide range of train velocities is discussed in
the next section.
Sustainable Bridges SB-4.7.2 2007-11-30 21 (46)
The values shown in Figures 4.8 and 4.9 correspond to when the train is located directly
above the back of the first abutment. Results are expressed in terms of net horizontal stress,
which represents the additional horizontal stress applied to the back of the structure due to the
train load, and maximum vertical deflections behind the abutment. The maximum net horizon-
tal stresses occur between depths of 0.2 and 0.45 m, and the maximum vertical deflections
occur approximately 0.6 m behind the train.
Sustainable Bridges SB-4.7.2 2007-11-30 22 (46)
80
Paxle = 125 kN
Paxle = 225 kN (freight)
40
Paxle = 250 kN
Paxle = 330 kN (freight)
20
0
0 100 200 300 400
Train Velocity (km/hr)
3
Maximum Vertical Deflection (mm)
2
Paxle = 125 kN
Paxle = 225 kN (freight)
Paxle = 250 kN
Paxle = 330 kN (freight)
1
0
0 100 200 300 400
Train Velocity (km/hr)
In general, as the train velocity increases, the net horizontal stresses in the ballast and subbal-
last layer increase. However, the relationship between train velocity and maximum net hori-
zontal stress on the back of the abutment is not linear. For the passenger trains, the largest
incremental increase in the maximum net horizontal stresses at the first bridge abutment oc-
curs when the train velocity increases from 250 to 350 km/hr.
Similarly, as the train velocity increases, the maximum vertical deflection behind the train in-
creases. The relationship between train velocity and maximum vertical deflection behind the
first abutment is also not linear; the largest increase in maximum vertical deflections occurs for
the case in which the train velocity increases from 250 to 350 km/hr.
Sustainable Bridges SB-4.7.2 2007-11-30 23 (46)
As can be seen in Figures 4.8 and 4.9, the non-linear relationships between train velocity
and maximum net horizontal stress and vertical deflections are more pronounced for the
heavier train axle load.
The distribution with depth of vertical and horizontal pressures caused by the passage of a
passenger train with an axle load of 250 kN at velocities of 50 and 350 km/hr are shown in
Figures 4.10 and 4.11, respectively. The following observations can be made based upon the
results shown in Figures 4.10 and 4.11:
Within the upper meter, both the vertical and horizontal stresses increase with increas-
ing train velocity; though the increase is relatively small for the vertical pressures. This
would indicate that the vertical pressures are affected more by train axle load, and the
horizontal pressures by train velocity. These trends can also be seen in Figure 4.5. 4.6.
As train velocity increases, the horizontal stresses increase. This can also be seen in
Figure 4.6 as a function of train axle load.
Below a depth of about 1.6 m, which is the bottom of the subballast layer, there is little
variation in horizontal stresses for both train velocities;
For a train velocity of 350 km/hr, negative net vertical and horizontal pressures are ob-
served around a depth of about 1.5 m. Negative values of pressure indicate that the
soil is in tension. The values of tension, however, are relatively small.
-1
-2
Depth (m)
-3
at bridge, v = 50 km/hr
at bridge, v = 350 km/hr
-4
-5
-0.4
-0.8
Depth (m)
-1.2
v = 50 km/hr
v = 350 km/hr
-1.6
-2
The Rayleigh wave velocity for each material is also provided in Table 4.2. The Rayleigh
wave velocity, VR, of a material may be estimated using the following expression (Hall 2000):
G (0.87 + 1.12 )
VR =
(1 + )
Sustainable Bridges SB-4.7.2 2007-11-30 25 (46)
E
Where G = shear modulus = , E = Elastic Modulus, = Poisson's ratio, and =
2 (1 + )
density.
The goal of this research is to evaluate the behaviour of bridge systems under high speed
trains, which may have velocities up to 350 km/hr in the future. The parametric analyses
described in the following sections were performed using a train velocity of 350 km/hr, and
axle loads of 125 and 250 kN. The variation in the maximum net horizontal stresses, net
h,max, with velocity and material stiffness values is given in Table 4.3. The values of maxi-
mum net horizontal stresses in Table 4.3 were observed between depths of 0.2 and 0.45 m,
and are for the case in which the train is directly above the back of the first abutment. The
values in Table 4.3 are averaged values over a distance of 1 m from the center line of the
abutment. The variation of net horizontal stresses with material property values is discussed
in the following sections.
The variation in the maximum vertical deflections, v,max, with train velocity and material stiff-
ness values is given in Table 4.4 for the case in which the train is above the back of the
abutment, and is discussed in the following sections.
Ballast stiffness
The elastic modulus of the ballast was varied over a range of 160 to 300 MPa. For an axle
load of 125 kN, changing the stiffness of the ballast material has a relatively small effect on
the net horizontal stresses behind the first bridge abutment (see Table 4.3). For an axle load
of 250 kN, as the stiffness of the ballast increases, the net horizontal stresses decrease. For
both axle loads, an increase in ballast stiffness leads to a decrease in maximum vertical de-
flections behind the bridge abutment.
Subballast Stiffness
The elastic modulus of the subballast was varied over a range of 47.9 to 300 MPa. When
the stiffness of the subballast layer is equal to 47.9 MPa, it is equal to the stiffness of the
embankment fill, and in this case, it is as if there is no subballast layer and the embankment
fill extends to the bottom of the ballast layer. When the stiffness of the subballast layer is
equal to 300 MPa, the subballast is 1.6 times stiffer than the ballast layer.
The stiffness of the subballast layer has a significant effect on the calculated maximum net
horizontal stresses, which occur at the ballast-subballast interface. For both axle loads, the
greatest increase in net horizontal stresses occurs for the case in which the softer subballast is
used. For this case, the Rayleigh wave velocity of the subballast and embankment fill is on
the order of 345 km/hr. Conversely, for both axle loads, as the stiffness of the subballast in-
creases, the maximum horizontal stresses decrease. Furthermore, for an axle load of 250
km/hr the decrease in horizontal stress due to a stiffer subballast layer is greater than the de-
crease in stress that is associated with decreasing the train velocity from 350 to 250 km/hr.
A cross-section of the net horizontal stresses at the first bridge abutment are shown in Figure
4.12 for the case in which the stiffest (E = 300 MPa) subballast layer was used and the train
velocity was equal to 350 km/hr. As can be seen in Figure 4.12, the maximum net horizontal
stresses occur at a depth of about 0.4 m, at the ballast-subballast layer interface. The net
horizontal stresses below a depth of 1 m are relatively small.
Sustainable Bridges SB-4.7.2 2007-11-30 27 (46)
Figure 4.12. Net horizontal stresses behind first bridge abutment (v=350 km/hr)
for stiff subballast case. (Contours given in kPa.)
Sustainable Bridges SB-4.7.2 2007-11-30 28 (46)
As can be seen in Table 4.4, as the stiffness of the subballast layer increases, vertical deflec-
tions behind the first abutment decrease.
Embankment Fill
The elastic modulus of the embankment fill was varied over a range of 20 to 70 MPa. For
both axle loads, an increase in the stiffness of the fill corresponds to a decrease in the net
horizontal stresses in the ballast and subballast layers. Furthermore, for both axle loads, the
decrease in horizontal stress due to a stiffer embankment fill layer is greater than the de-
crease in stress that is associated with decreasing the train velocity from 350 to 250 km/hr.
A cross-section of the net horizontal stresses at the first bridge abutment are shown in Figure
3.13 for the case in which the softest fill layer was used and the train velocity was equal to
350 km/hr. For the heavier axle load, a decrease in stiffness of the embankment fill has little
effect on the calculated magnitude of maximum net horizontal stress. As can be seen in Fig-
ure 4.13, the maximum net horizontal stress occurs at about a depth of 0.2 m and is ap-
proximately 76 kPa. However, net horizontal stresses on the order of 10 kPa are observed to
a depth of approximately 3.8 m. The maximum increase in stresses in the embankment fill,
however, occurs slightly ahead of the train. Net horizontal stresses on the order of 30 kPa
were observed at the subballast-fill interface slightly ahead of the moving train for this case.
For both axle loads, the greatest vertical deflection behind the first bridge abutment was cal-
culated for the case in which the soft embankment fill was used. This high value of calculated
deflection occurs approximately 3 m behind the train.
Sustainable Bridges SB-4.7.2 2007-11-30 29 (46)
Figure 4.13. Net horizontal stresses behind first bridge abutment (v=350 km/hr)
for soft fill case. (Contours given in kPa.)
Sustainable Bridges SB-4.7.2 2007-11-30 30 (46)
The greatest increase in vertical and horizontal stresses occur in the ballast and sub-
ballast layers; these increases occur approximately beneath the train when the train is
above the back of the abutment.
The greatest increases in vertical and horizontal stresses are concentrated within the
upper 1.0 to 1.2 m of the system. The vertical and horizontal stresses caused by the
passing of a train decrease with depth. Similar observations have been made in field
tests by Kolisoja et al. (2000).
As train axle loads increase, the vertical and horizontal stresses, and vertical deflec-
tions beneath the train increase. Similar observations have been made by Katzenbach
and Ittershagen (2005) and Hunt (1997), respectively.
Within the upper meter, both the vertical and horizontal stresses increase with increas-
ing train velocity; though the increase in vertical pressure is relatively small. The verti-
cal pressures are affected more by train axle load, and the horizontal pressures by train
velocity. Additionally, there is a substantial increase in both the maximum horizontal
stresses and the vertical deflections when the train velocity increases from 250 to 350
km/hr.
As the stiffness of the ballast and/or subballast materials increase, net horizontal
stresses exerted on the bridge abutment decrease. Horizontal stresses in the em-
bankment fill are relatively small, but increase when a soft fill is used.
Changing the stiffness of upper ballast material has little effect on the calculated verti-
cal deflections behind the first bridge abutment. This is probably due to the fact that
the ballast layer is relatively thin.
As the stiffness of the subballast layer increases, vertical deflections behind the first
abutment decrease. For the majority of the cases evaluated, when the train is located
directly above the back of the first bridge abutment, the maximum vertical deflections
are observed to occur approximately 1 m behind the train, except for the case in which
a soft fill is used. When a soft embankment fill is used, vertical deflections increase
substantially, and these deflections are observed to occur 3 m behind the train, as if
there is a stern wave behind the train. The presence of a wave-like phenomenon in
front of and in back of the train has also been documented by Adolfsson et al. (1999b)
and Kopf and Adam (2005).
Sustainable Bridges SB-4.7.2 2007-11-30 31 (46)
Like the concrete bridge geometry, the masonry arch bridge geometry has the embankment
profile that is shown in Figure 5.2. The profile consists of an upper ballast layer 0.4 m thick,
which is underlain by a layer of subballast 1.2 m thick, which in turn is underlain by embank-
ment backfill. A stiff, non-yielding, bearing layer is assumed to exist beneath the embank-
ment fill. As can be seen in Figure 5.2, the profile is symmetrical about its centre-line.
7.4 m
3.3 m Train Load
2.5 m
Ballast
0.4 m
Subballast 2 1.2 m
1
Embankment Fill
3.8 m
The material property values used in the numerical analyses are listed in Table 5.1. A linear-
elastic perfectly plastic model with a Mohr-Coulomb failure criterion was used to represent the
ballast, subballast, and embankment fill.
No structural elements were used to model the bridge, or the rail and sleepers. It was desired
to model the train dynamically, and since the timestep in the dynamic mode is determined by
the largest material stiffness, using structural elements would require unrealistically long
model run times, and may cause numerical problems. Therefore, the bridge structure was
represented by fixed gridpoints at the bridge-soil interface.
A series of numerical analyses were also performed using the masonry arch bridge geometry
shown in Figure 5.1. Three train load and velocity configurations were considered:
1. The high speed X2000 train with an axle load of 125 kN, axle spacing of 2.9 m, and
velocity of 350 km/hr,
2. A freight train with an axle load of 225 kN, axle spacing of 1.8 m, and velocity of 100
km/hr, and
3. A freight train with an axle load of 330 kN, axle spacing of 1.8 m, and velocity of 100
km/hr.
Figure 5.4. Net stresses in arch bridge under passing of high speed X2000 train,
Paxle = 125 kN, v = 350 km/hr (contours in kPa)
Sustainable Bridges SB-4.7.2 2007-11-30 34 (46)
350 350
300 300
250 250
200 200
150 150
100 100
50 50
150 200 250 300 150 200 250 300
Figure 5.7. Arch bridge geometry: projected contours of net stress at a depth of 0.5 m
as a function of train velocity and axle load
Projected contours of net horizontal and net vertical stress at the point of the arch at a
depth of 0.5 were developed. Plots of the contours are shown in Figure 5.7. As can be seen
in Figure 5.7, at a depth of about 0.5 m, net horizontal stresses increase with increasing train
velocity and axle load; however increasing the train velocity over a range of about 50 to 150
km/hr has little affect on the horizontal stresses for a given axle load. At a depth of 0.5 m,
the vertical stresses increase with axle load and are not affected significantly by increases in
train velocity.
Sustainable Bridges SB-4.7.2 2007-11-30 36 (46)
4
Normalized Vertical Stress
6
10
12
at crown
14 at 3/4 point
16
4
Normalized Vertical Stress
10
12
at crown
14 at 3/4 point
16
Figure 5.8. Influence lines at the crown and point of arch bridge (Axle Load = 225 kN)
Sustainable Bridges SB-4.7.2 2007-11-30 37 (46)
4
6
Normalized Vertical Stress
8
10
12
14
16
18
20 at crown
at 3/4 point
22
24
4
6
Normalized Vertical Stress
10
12
14
16
18
20 at crown
at 3/4 point
22
24
Figure 5.9. Influence lines at the crown and point of arch bridge (Axle Load = 330 kN)
Sustainable Bridges SB-4.7.2 2007-11-30 38 (46)
Influence lines, in the form of normalized vertical stress against time, are shown in Figures
5.8 and 5.9 for trains with axle loads of 225 and 330 kN, respectively. The influence lines
shown in Figures 5.8 and 5.9 pertain to points directly above the crown and the point of
the arch. As an example, under the passing of a train with an axle load of 225 kN and a ve-
locity of 100 km/hr, the arch structure will experience a relative increase in vertical stress at
the crown on the order of about 16 times the initial stress.
The normalized vertical stress in Figures 5.8 and 5.9 is defined as the vertical stress induced
by the passing train divided by the initial (static) vertical stress. It should be noted that the
initial vertical stresses at the and points are higher than that at the crown. Therefore,
even though the crown feels a greater relative increase in vertical stress, the total vertical
stresses induced at the crown and the and points are similar in total magnitude. As has
been demonstrated previously, the induced vertical stresses decrease with depth; in most
cases, most of the induced vertical stresses due to a passing train are concentrated within
the upper 1.0 to 1.5 m.
There are a few general observations that can be made based upon Figures 5.8 and 5.9.
First, as would be expected, the relative vertical stresses induced on the arch structure in-
crease with train axle load. However, vertical stresses are not greatly affected by train veloc-
ity. Second, since there is less soil above the crown of the arch to dissipate stresses, the
crown of the arch will experience the greatest spike in vertical stress under the passing of a
train.
The greatest increase in vertical and horizontal stresses occur in the ballast and sub-
ballast layers; these increases occur approximately beneath the train when the train is
above the bridge.
As train axle loads increase, the vertical and horizontal stresses, and vertical deflec-
tions beneath the train increase.
Horizontal stresses increase with increasing train velocity and axle load. The vertical
pressures are not significantly affected by train velocity for the range of velocities
evaluated. However, vertical pressures may be expected to increase at very high train
velocities (i.e. train velocities that approach the critical velocity of the bridge and em-
bankment system.)
The geometry of the arch bridge is different than the concrete bridge in two significant ways:
(1) for masonry arch bridges there is a gradual change in stiffness, while for a concrete
bridge the change is more abrupt, and (2) for the masonry arch structure, the ballast and
subballast are confined within a relatively small volume. Due to the latter factor, additional
vertical and horizontal stresses induced by the train are distributed immediately to the struc-
ture within the arch. A few additional general observations can be made based upon the
analyses using the masonry arch bridge:
Sustainable Bridges SB-4.7.2 2007-11-30 39 (46)
As the train axle load increases, the vertical and horizontal stresses below the train in-
crease, as expected. Significant vertical stresses are exerted on the top of the arch.
These stresses are concentrated primarily beneath the width of the train.
For a passenger train with an axle load of 125 kN, the maximum net horizontal stresses
occur to depths of about 0.7 m; however, for the freight trains with axle loads of 225
and 330 kN, net horizontal stresses on the order of 15 and 20 kPa, respectively, are
observed down to depths of 0.9 to 1.0 m.
Sustainable Bridges SB-4.7.2 2007-11-30 40 (46)
6 LIMITATIONS
The scope of this work included performing numerical analyses to evaluate the load distribu-
tion and deflections in bridge transition zones due to the passing of trains. No field or labora-
tory tests were performed. This document is a background document, intended to provide
insight into the general behavior of bridge systems under the passing of high speed trains.
No design methods or recommendations are presented herein.
The results presented in this document are only for the conditions investigated. The findings
of the numerical analyses should be verified against laboratory scaled tests and/or full-scale
field tests.
Sustainable Bridges SB-4.7.2 2007-11-30 41 (46)
7 CONCLUSIONS
The scope of this work included performing numerical analyses to evaluate the load distribu-
tion and deflections in bridge transition zones due to the passing of trains. No field or labora-
tory tests were performed. Numerical analyses consisted of three-dimensional dynamic
analyses. Using a bridge geometry representative of a typical European concrete bridge, a
numerical parametric study was performed to study the effects of 1.) the train load, 2.) the
train velocity, and 3.) the stiffness of the ballast, subballast and embankment fill. Numerical
analyses were also performed using a bridge geometry representative of European masonry
arch bridges.
The greatest increases in vertical and horizontal stresses are concentrated within the
upper 1.0 to 1.2 m of the system. The additional vertical and horizontal stresses
caused by the passing of a train decrease with depth.
As train axle loads increase, the vertical and horizontal stresses, and vertical deflec-
tions beneath the train increase.
As train velocity increases, the horizontal stresses and vertical deflections increase,
however, they do not increase linearly with increasing velocity. As the velocity of the
train approaches the Rayleigh wave velocity of the embankment system, significant
horizontal stresses and vertical deflections develop.
As the stiffness of the ballast and/or subballast materials increase, net horizontal
stresses exerted on the bridge decrease. Horizontal stresses in the embankment fill
are relatively small, but increase when a soft fill is used.
When a soft embankment fill is used, vertical deflections increase substantially, and
these deflections are observed to occur 3 m behind the train, as if there is a stern
wave behind the train.
The numerical methods used in this research were verified against data from an instru-
mented test embankment published by the Finnish railway administration. However, the Fin-
nish test embankment did not include a bridge structure. The results presented in this report
are intended to provide insight into the behavior of railway bridge transition zones. The re-
sults presented in this report are valid for the conditions studied; however, field tests and/or
laboratory tests are needed to verify the numerical analyses.
Sustainable Bridges SB-4.7.2 2007-11-30 42 (46)
8 REFERENCES
Adolfsson, K., Andrasson, B., Bengtsson, P., and Zackrisson, P. (1999a). High speed train
X2000 on soft organic clay measurements in Sweden. Proceedings of the European
Conference on Soil Mechanics and Geotechnical Engineering, Geotechnical engineering
for transportation infrastructure, Amsterdam, June 1999. Vol. 3: 1713 1718.
Adolfsson, K., Andrasson, B., Bengtsson, P-E., Bodare, A., Madshus, C., Massarsch, R.,
Wallmark, G., & Zackrisson, P. (1999b). Evaluation and Analyses of Measurements from
the West Coast Line. Swedish Geotechnical Institute Report.
Chen, Y.M., Wang, C.J., Chen, Y.P., and Zhu, B. (2005). Characteristics of stresses and set-
tlement of ground induced by train. Environmental Vibrations, Takemiya (ed.) Taylor and
Francis Group, London: 33 42.
Cojocaru, E.C., Irschik, H., and Schlacher, K. (2003). Concentrations of Pressure Between
an Elastically Supported Beam and a Moving Timoshenko-Beam. Journal of Engineering
Mechanics, 129, no. 9, 1076-1082.
ERRI (European Rail Research Institute) (1999). State of the Art Report, Bridge Ends, Em-
bankment Structure Transition. ERRI Report D 230.1/RP3, Utrecht, 100 pp.
Hall, L. and Bodare, A. (1999). Prediction of frequency content in train induced ground vibra-
tions based on a beam-on-Winkler-foundation model. Proceedings, Geotechnical Engi-
neering for Transportation Infrastructure, Barends et al. (eds), Balkema, Rotterdam: 1803
- 1808.
Hall (2000). Simulations and Analyses of Train-Induced Ground Vibrations. Doctoral Thesis
1034, Division of Soil and Rock Mechanics, Department of Civil and Environmental Engi-
neering, Royal Institute of Technology, Stockholm.
Hunt, H.E.M. (1997). Settlement of railway track near bridge abutments. Proceedings Inst.
Civil Engs., Transp., 123, no. 1: 68-73.
Itasca (2005). FLAC3D Fast Lagrangian Analysis of Continua in 3 Dimensions, Users Guide,
Itasca Consulting Group, Inc., Minneapolis, Minnesota, USA.
Kaynia, A.M., Madshus, C., and Zackrisson, P. (2000). Ground Vibrations From High-Speed
Trains: Prediction and Countermeasure. Journal of Geotechnical and Geoenvironmental
Engineering, 126, no. 6, 531-537.
Kolisoja, P., Jrvenp, I., Mkel, E., and Levomki, M. (2000) Instrumentation and model-
ling of track structure, 250 kN and 300 kN axle loads, Finnish Rail Administration, Publica-
tion A 10/2000, 99 pp (in Finnish).
Kopf, F. and Adam, D. (2005). Dynamic effects due to moving loads on tracks for high-speed
railways and on tracks for metro lines. Proceedings of the 16th International Conference
on Soil Mechanics and Geotechnical Engineering, Osaka, September, 2005. Vol. 3: 1735
1740.
Li, D. and Davis, D. (2005). Transition of Railroad Bridge Approaches. Journal of Geotechni-
cal and Geoenvironmental Engineering, 131, no. 11, 1392-1398.
Lundqvist, A. and Dahlberg, T. (2004). Dynamic train/track interaction including model for
track settlement evolvement. Progress report, March 2, The SUPERTRACK project,
Linkping University, Sweden.
Madshus, C., Lacasse, S., Kaynia, A. and Hrvik, L. (2004). Geodynamic Challenges in High
Speed Railway Projects. Geotechnical Engineering for Transportation Projects, ASCE
GSP 126, pp. 192-215.
PRIORITY 6
SUSTAINABLE DEVELOPMENT
GLOBAL CHANGE & ECOSYSTEMS
INTEGRATED PROJECT
Sustainable Bridges SB-4.7.3 2007-11-30 2 (71)
This report is one of the deliverables from the Integrated Research Project Sustainable Bridges - Assessment for
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework Pro-
gramme. The Project aims to help European railways to meet increasing transportation demands, which can only
be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
passenger trains. This requires that the existing bridges within the network have to be upgraded without causing
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and econ-
omy of the railways.
A consortium, consisting of 32 partners drawn from railway bridge owners, consultants, contractors, research
institutes and universities, has carried out the Project, which has a gross budget of more than 10 million Euros.
The European Commission has provided substantial funding, with the balancing funding has been coming from
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule Tech-
nical University has undertaken the scientific leadership.
The Project has developed improved procedures and methods for inspection, testing, monitoring and condition
assessment, of railway bridges. Furthermore, it has developed advanced methodologies for assessing the safe
carrying capacity of bridges and better engineering solutions for repair and strengthening of bridges that are found
to be in need of attention.
The authors of this report have used their best endeavours to ensure that the information presented here is of the
highest quality. However, no liability can be accepted by the authors for any loss caused by its use.
Table of Contents
Summary .................................................................................................................... 5
1 Introduction............................................................................................................. 6
1.1 Subject of the study ....................................................................................... 6
1.2 Practical application of the study.................................................................... 6
1.3 Applied terminology ....................................................................................... 6
1.4 Applied notation ............................................................................................. 8
2 Taxonomy of defects in masonry arch bridges ....................................................... 9
2.1 General conception........................................................................................ 9
2.2 Hierarchical classification of masonry bridge defects..................................... 9
2.3 Defects and structure condition.................................................................... 11
3 Methods applied in analysis of damaged masonry arch bridges .......................... 12
3.1 Classification of methods for analysis of damaged masonry arch bridges... 12
3.2 Analytical methods....................................................................................... 14
3.2.1 Introduction ....................................................................................... 14
3.2.2 The maximum stress analyses.......................................................... 14
3.2.3 The thrust line analyses .................................................................... 18
3.2.4 The mechanism methods.................................................................. 22
3.3 Computer-based applications ...................................................................... 26
3.3.1 Introduction ....................................................................................... 26
3.3.2 Archie-M ........................................................................................... 26
3.3.3 RING................................................................................................. 28
3.3.4 FEM systems .................................................................................... 30
3.3.5 DEM systems.................................................................................... 42
3.3.6 DDA systems .................................................................................... 43
3.3.7 Explicit formula analysis.................................................................... 45
3.4 Semi-empirical methods .............................................................................. 45
3.4.1 Introduction ....................................................................................... 45
3.4.2 The MEXE method............................................................................ 46
3.5 Comparison of the analysis methods ........................................................... 48
4 Conception of advanced analysis of damaged masonry arch bridges by means of
FEM ..................................................................................................................... 50
4.1 General idea ................................................................................................ 50
4.2 Selection of defects types ........................................................................... 50
4.3 Parameters of defects description............................................................... 50
Sustainable Bridges SB-4.7.3 2007-11-30 4 (71)
Summary
The present study brings up an issue of defects in masonry arch bridges and possi-
bilities of their analysis by means of deterministic analytical methods. Description of
the defects is based on the taxonomy developed within WP3 of this project. Pre-
sented methods used in analysis of masonry arch bridges with defects are classified
in three groups: analytical methods, computer-based applications and semi-empirical
methods. The analytical methods are based on: the maximum stresses, the thrust
line or the mechanism method. For each method detailed description from the point
of view of the defects modelling possibilities is provided. Ways of representation for
various defect types in all the methods is explained and their usefulness is evaluated.
Precise analysis of selected defects influence on the ultimate load for various bridge
geometries is presented with application of the Finite Element Method. Detailed de-
scription of the proposed 2D model of the bridge is included in the report. Methods of
modelling for selected defects in FEM is explained. Some examples of the analyses
and results of the defects influence on the ultimate load are presented in several
diagrams.
General conclusions on application of all presented methods of the defects model-
ling are given. The influence of the main defect types on masonry bridge structures
analysed by means of FEM is commented.
Sustainable Bridges SB-4.7.3 2007-11-30 6 (71)
1 Introduction
1 Formal information
This technical report is prepared on the basis of Contract No. TIP3-CT-2003-001653
between the European Community represented by the Commission of the European
Communities and the Skanska Teknik AB contractor acting as Coordinator of the
Consortium.
Presented report is a contribution of Wrocaw University of Technology to the Guide-
line SB-LRA (2007).
CONTAMINATION
DEFORMATION
DETERIORATION
DISCONTINUITY
DISPLACEMENT
LOSS OF MATERIAL
Figure 2.2 Main types of masonry bridge defects according to SB-CAI (2007)
Sustainable Bridges SB-4.7.3 2007-11-30 10 (71)
Definitions of all defect types and their examples observed in stone and brick bridges
are given in Table 2.1.
Table 2.1: Classification of main types of defects observed on masonry arch bridges
Type of Examples
defect Stone bridge Brick bridge
CONTAMINATION
Appearance of any type
of a dirtiness
or a plant vegetation
DEFORMATION
Geometry changes incompatible
with the project, with changes
of mutual distances
of structure points
DETERIORATION
Degradation of physical
and chemical structural features
DISCONTINUITY
Break in a structure
material continuity
DISPLACEMENT
Displacements of a structure or
its part incompatible with project
but without changes of mutual
distances of structure points
LOSS OF MATERIAL
Decrease of structure
material amount
Sustainable Bridges SB-4.7.3 2007-11-30 11 (71)
In this study only the types of defects influencing the load carrying capacity are being
considered.
Sustainable Bridges SB-4.7.3 2007-11-30 12 (71)
Analytical methods
Computer-based
applications Maximum stress The thrust line The mechanism
analysis analysis method
Archie-M
RING
FEM systems
DEM systems
DDA systems
Explicit formulae
Another approach to analysis of the load carrying capacity of masonry bridges using
theoretical bases comprise semi-empirical methods. However these methods are
also based on experimental techniques like laboratory tests or field tests. They usu-
ally have a form of simple formulae.
An example of a semi-empirical method is:
The MEXE method.
Figure 3.2 Geometry models in maximum stress analyses: 3-pin (a, b), 2-pin (c)
and fixed-end arch rib (d) or spatial frame (e)
Material models:
1. Elastic
Main features:
constitutive model: linear elastic,
homogenous,
required material properties: E, fc, ft, ,
2. Inelastic
Main features:
constitutive model: linear elastic-perfectly brittle or linear elastic-perfectly
plastic, no tensile resistant (NTR),
homogenous,
required material properties: E, fc, ,
Sustainable Bridges SB-4.7.3 2007-11-30 16 (71)
M P (s )y (s )
EI (s )
ds
H= eq. 1
y (s )
2
EI (s )
ds
As the calculation result the internal forces from applied loads and hence stresses
are obtained. Then the stresses are compared with the remaining material resistance
and in this way the load carrying capacity can be estimated.
An idea of considering plastic hinge formation (especially in fixed-end arch scheme)
is introduction of pins in places where highest stresses are detected.
Figure 3.4 Example of distribution of the effective thickness of the arch barrel in the
inelastic method (Brencich et al. 2001)
Figure 3.5 The force network for an arch barrel for the self-weight and a point load
(O'Dwyer 1999)
Material models:
1. Elastic
Main features:
constitutive model: linear elastic, infinite compression strength,
homogenous,
required material properties: E, ,
2. Elastic, NTR
Main features:
constitutive model: linear elastic, no tensile resistant (NTR), infinite
compression strength,
homogenous,
required material properties: E, ,
1. Elastic-plastic
Main features:
constitutive model: non-linear elastic-plastic, no tensile resistant (NTR),
homogenous,
required material properties: E, fc, ,
is equal one-third of the total depth which is given by eq. 2. This criterion has been
arrived at on the basis of elastic theory assuming elastic material.
M d
e= eq. 2
N 6
Such a rigorous limit makes the method extremely conservative and at the same time
is very difficult to be met. It can be satisfied only in cases when the structure has
been properly designed and the dead loads considerably dominates over live loads.
The above-mentioned problem caused that the method was relaxed to the middle
half rule what increased the size of allowed core for the force resultant position to the
half of the cross-section (eq. 3).
M d
e= eq. 3
N 4
The last variant of the method was proposed by Heyman (1982). He concluded that
even if the line of thrust runs outside the allowed core in one section it does not
threaten the safety of the whole structure. A structure will collapse only if the line of
thrust reaches the edges of the arch barrel at least in four sections that would convert
it into mechanism. According to this idea the allowed space for the thrust line is the
whole section of the arch barrel what can be written as:
M d
e= eq. 4
N 2
Hence the resultant force in every section of the arch can not lie nearer to the edge
of the arch then t/2 (Figure 3.6).
force resultant
t
e
yield d
t block
zone of thrust
Figure 3.7 Application of the mechanism method for assumed hinge location
A simplified approach uses an assumed mode of a mechanism where position of
hinges is established as the most probable one. For a concentrated load pattern as in
Figure 3.7 one of the hinges should be introduced under the force (in point B). Re-
maining hinges can be located at the springings (A, D) and in the place providing
creation of a geometrically variable scheme (C). Looking for the best mechanism
mode few various hinge locations should be taken until the minimum load is found.
The solution for a given scheme can be carried out by taking bending moments about
hinge points equal to zero or solving equation of virtual work for all parts of the arch
barrel with self-weights Gi and geometrically imposed displacements i (Figure 3.7,
eq. 8).
a) b)
Figure 3.8 Geometry models in the mechanism method analyses: plane arch (a) and
spatial vault (b)
Material models:
1. Rigid
Main features:
constitutive model: infinite compression strength, no tensile resistant (NTR),
homogenous,
required material properties: ,
2. Rigid-plastic
Main features:
constitutive model: rigid-perfectly plastic, no tensile resistant (NTR),
homogenous,
required material properties: fc, ,
a) b) c)
Figure 3.10 Feasible relative movements of adjacent blocks: rotation about extrados
(a), about intrados (b) or sliding (c)
Yield
Plastic block
Hinge
on the section width B what means that modelling of parallel cracks and fractures or
losses by reduction of B is realized through modification of t.
Defects possible to consider are presented in Table 3.4.
Table 3.4 Defect modelling in the mechanism methods
Defect Representation Comment
Applicable only to the whole cross-
Modification of material properties section;
Deterioration (fc) Not applicable in rigid blocks
(strength method version
reduction)
Modification of a plastic hinge location Not applicable in rigid blocks
from the arch barrel edge (t/2) method version
Discontinuity Modification of the arch barrel width Applicable only to the whole cross-
(longitudinal (B) section
crack and Modification of a plastic hinge location
fracture) from the arch barrel edge (t/2)
Discontinuity Modification of number of rings Simulates separation of rings;
(delamination) (n) Applicable to the whole arch barrel
Reduction of the arch barrel thickness Applicable only to the whole cross-
Loss of material (d) section;
(brick/stone) Modification of a plastic hinge location
from the arch barrel edge (t/2)
Deformation Based on modification of block ge-
Modification of static scheme geometry
(deflection) ometry without creation of hinges
Displacement
(translation and Enforced movements of supports
rotation)
3.3.2 Archie-M
Calculation are carried out on a static scheme of a 3-pin arch where locations of the
pins is chosen as the most likely one for a given load pattern. In respect of that the
program presents a response of the structure to any value of loads in the form of the
most probable zone of thrust location in the profile of the arch (Figure 3.12). It also
graphically indicates location of the hinges. The condition for the structure to be in
equilibrium is that the zone of thrust exists in the interior of the cross sections along
the whole arch. The limit state giving the load carrying capacity can be obtained in-
creasing the live load value up to the moment when the zone of thrust begins touch-
ing the edge of the arch profile in the fourth point.
dm
d
d
3.3.3 RING
2. Elastic, NTR
Main features:
constitutive model: linear elastic, no tensile resistant (NTR), infinite
compression strength,
homogenous,
required material properties: E,
3. Elastic-plastic
Main features:
constitutive model: non-linear elastic-plastic, limited compressive and ten-
sile strength,
inhomogeneous,
required material properties: (), fc, ft, ,
4. Elastic-plastic, NTR
Main features:
constitutive model: linear elastic-perfectly plastic, no tensile resistant (NTR),
finite compression strength,
homogenous,
required material properties: E, fc,
5. Elastic-brittle, NTR
Main features:
constitutive model: linear elastic-perfectly brittle, no tensile resistant (NTR),
finite compression strength,
homogenous,
required material properties: E, fc,
3.3.4.3 1D modelling
There are various 1D modelling techniques dependent on the applied material model
and related analysis algorithm.
The simplest approach is based on elastic material model and comprise the numeri-
cal solution of the elastic maximum stress analysis presented in chapter 3.2.1. How-
ever the most commonly applied model is a fixed-end arch rib. Pinned models which
impose zero bending moments in pins are usually neglected in that kind of analysis.
The arch is modelled as a set of one-dimensional straight beam elements with re-
quired geometric and material properties. In the analysis the internal forces and
stresses are calculated and compared than with the material strength values.
Sustainable Bridges SB-4.7.3 2007-11-30 32 (71)
1
4
2 6
7 5
3
Figure 3.17 Elastic-plastic analysis using 1D modelling in FEM system
Sustainable Bridges SB-4.7.3 2007-11-30 33 (71)
A slightly different approach was presented by Choo et al. (1991) using tapered
beam elements and elastic-brittle material. Here both cracked in tension and crushed
in compression areas of the arch barrel section are removed for calculation in the
next step of an incremental procedure. The axis of an updated arch geometry is dis-
placed to the central line of the current effective arch (Figure 3.18). The effective
thickness is calculated in every node and varies linearly between them on each sec-
tion length.
3.3.4.5 2D modelling
Two-dimensional model of a masonry arch bridge is usually created as a vertically
situated plane arch representing a profile of a unit width of the arch barrel. There are
several modelling techniques with respect to the homogeneity of the arch material.
Figure 3.19 shows the applied methods differing from each other in the kind and
boundaries of the employed materials. In the first approach (a) the whole arch barrel
is modelled with a homogenous material what is called a macro-modelling. Another
methods (b) and (c) a simplified micro-modelling include a division into discrete
segments acting between one another through the contact surfaces. It enables de-
tachment of adjacent parts simulating cracking. The next techniques (d)-(f) incorpo-
rate additional joint elements of a different material properties modelling selected
joints of a masonry what can be called a detailed micro-modelling. Depending on
the required level of accuracy and the expected failure mode there are modelled ra-
dial joints only (d), tangential joints only (e) or both of them (f). In the approaches
predicting division into masonry blocks these segments include a piece of masonry
consisted of few units. Such a method is detailed enough to give reasonable results
with regard to a failure mode and the load carrying capacity value in comparison with
the experimental tests. More fine mesh e.g. responding to individual masonry units,
does not bring much benefits but significantly complicates and prolongs the solution.
Figure 3.22 Post failure uniaxial behaviour of concrete like material: perfectly brittle
(a), perfectly plastic (b) or strain softening (c) (Loo and Yang 1991)
The perfectly brittle model is practically inaccessible due to difficulties in solution
convergence reaching the limit envelope. On the other hand the perfectly plastic ap-
proach, the most easily attainable one from the numerical point of view, leads to
overestimation of the actual global limit load for the structure but is sufficient in de-
termining the failure mode. The most truthful imitation of the real masonry material
provides the model employing the strain softening law. It is appropriate especially in
tension but can be also applied in compression. Stresses after reaching the critical
value cr are gradually decreasing down to total loss of the material strength (Figure
3.22(c)). The slope of the descending branch should be taken so to give the maxi-
mum strain about 36 times of the critical value (k = 36). The softening function in
the form of -w relationship where w is a crack opening can be linear or nonlinear
providing a constant area under the -w curve being equal to dissipated fracture en-
ergy Gf (Hillerborg 1976).
a)
b)
3.3.4.7 3D modelling
The most advanced but at the same time the most demanding approach to analysis
of masonry arch bridges provides three-dimensional modelling in FEM. The main ad-
vantage of this method is an opportunity to model a whole structure including span-
drel walls and/or abutment wings. With respect to such a detailed computer model
also a scope of accessible results is more extensive and precise. Contrary to the
simpler models it is possible to consider specific structural behaviour of a bridge in-
volving transverse effects and analyse different or more localized failure types. A new
feature of such a technique is the transverse dimension of the model enabling taking
into account both various (not only uniform) configuration of loads and variable re-
sponse of the structure in this transverse direction like variable stress distribution.
Sustainable Bridges SB-4.7.3 2007-11-30 40 (71)
Figure 3.28 Model of half of a symmetric masonry bridge structure (Fanning &
Boothby 2001)
In comparison with 2D models here the dominant modelling technique is based on
the homogenized composite material not dividing an arch barrel into masonry blocks.
The material behaviour in pre- and post failure states is described in an analogical
way as in 2D models. An obvious difference concerns definition of a limit domain in a
three-dimensional space. The most simplified idea is given by Rankine type criterion
but more refined and more appropriate for concrete-like materials limit surface is de-
termined by Willam-Warnke criterion (Figure 3.29).
Sustainable Bridges SB-4.7.3 2007-11-30 41 (71)
a) b)
Figure 3.30 Accessible outcome of 3D analysis in FEM: stress distribution (a), defor-
mation (b) (Frunzio et al. 2001, Fanning & Boothby 2001)
In the Discrete Element Method the structural analysis is based on explicit algorithms
related to solving the dynamic equations of equilibrium between blocks. An important
fact which has to be taken into account here is that the geometry of the system as
well as the number and type of contacts between the blocks, may change during the
analysis.
An output of such an analysis encompasses presentation of stresses distribution,
progressive failure mode associated with cracks propagation and maximal load pos-
sible to apply to the structure.
Discontinuity
Fcm = 0,30,7 Diagonal cracks
(skew)
Reduction of the load carrying ca-
For a 4-ring arch barrel, where N is a number
pacity due to ring separation by ap- of separated rings
Discontinuity
plying coefficients: R4-rings = 1 0,2n
(delamination)
For a 6-ring arch barrel, where N is a number
R6-rings = 1 0,146N of separated rings
Theoretical bases
Maximum stress analyses
Elastic arch rib z z
Inelastic arch rib z z z
Elastic spatial frame z z z
The thrust line analyses
SB-4.7.3
Line of thrust z z z
Zone of thrust z z z
Funicular network z z z
The mechanism methods
Rigid blocks z z z
Rigid-plastic blocks z z z
Volumetric blocks z z z z
Computer-based applications
Archie-M z
RING
ver. 1.5 z
ver. 2.0 z z z z
FEM
1D z z
2D z z z z z z z z
3D z z z z z z z
2007-11-30
DEM z z z
DDA z z z
Explicit formulae
Semi-empirical method
MEXE
49 (71)
4.3.2 Location ()
Description applied for deterioration, discontinuity and losses:
- value of the ratio between the longitudinal coordinate indicating the central
point of a defect xc to the whole arch barrel length l (Figure 4.1), (eq. 14):
x cd ,c,l
d,c,l = eq. 14
l
4.3.3 Extent ()
Description applied for deterioration, discontinuity and losses:
- value of the ratio between a defect length l to the whole arch barrel length l
(Figure 4.1), given on the percentage basis (eq. 15):
l d ,c,l
d,c,l = 100% eq. 15
l
4.3.4 Intensity ()
Description applied for deterioration, discontinuity and losses:
- value of the ratio between a damaged height in the cross-section d to the
designed arch barrel thickness d (Figure 4.1), given on the percentage basis
(eq. 16):
d d , c, l
d,c,l = 100% eq. 16
d
et al. 2003). In the study numerical examples of defects incorporation are presented
on the most common constructions among masonry arch bridges: single-span, rec-
tangular in plane structure with single-ring and segmental arch barrel. However the
same techniques can be applied to any structural types of masonry arch bridges
modelled in the proposed way.
what provides good representation of a real structure behaviour. Besides the number
of the blocks has very limited influence on the load carrying capacity what was ana-
lysed by authors. Thickness of joint elements is 1 cm which comprises an average
dimension of real mortar joints. The densities of both masonry blocks b and mortar j
are the same and are equal to 20 kN/m3.
block elements
joint elements
Model of the arch barrel geometry is created of 4-node quadrilateral plane elements
with 2 degrees of freedom in each node. Recommended ratio between element sides
is near to 1, therefore the number of elements on the arch barrel thickness should be
large enough.
Both ends of the arch barrel composed of joint elements are supported by fixed
nodes. It models connection of the arch springings with abutments which are as-
sumed to be rigid.
Considered constitutive models of material are different for masonry blocks and
joints. For simplification of a numerical problem the material of masonry blocks is as-
sumed to be linear-elastic with unconstrained compressive and tensile strengths de-
fined only by modulus of elasticity Eu and Poisons ratio u. Values of these properties
are taken as average values for masonry prism i.e. 10 GPa and 0,16 respectively.
More complex constitutive model of joint elements represents a concrete-like material
behaviour. Thus it is non-linear in compression with compressive strength fc equal 10
MPa and linear in tension with considerably limited tensile strength ft about 100 kPa
(Figure 4.5).
Sustainable Bridges SB-4.7.3 2007-11-30 54 (71)
4.4.3 Backfill
Model of the backfill has to be created carefully because it comprises very important
and fully-fledged structural component. It fulfils a dual role in the structural behaviour
of a masonry bridge: provides transfer of loads from the track structure to the arch
barrel and exerts a stabilising influence on deformed arch barrel through a mobilised
passive earth pressure.
Proposed element type for modelling the backfill is a 3-node triangular plane with 2
degrees of freedom in each node.
The applied properties of the soil are typical for most encountered structures. Their
values are as follows: modulus of elasticity Es = 100 MPa, Poisons ratio s = 0,3,
angle of internal friction = 45 and cohesion c = 30 kPa. Limit domain is defined by
Drucker-Prager criterion commonly applied to granular media. The density of the soil
is s = 18 kN/m3.
Sustainable Bridges SB-4.7.3 2007-11-30 55 (71)
One of the most important features of the used backfill model is its interaction with
the arch barrel. It is realized by introducing a contact surface between these two
parts of the structure. This interface provides mutual reaction of both contacting ele-
ments, disables their overlapping and enables total separation. Possible relative
transverse displacements of the arch barrel and the backfill is ruled by friction defined
by a value of a coefficient of friction = 0,5.
In numerical examples assumed height of the backfill over the arch barrel crown is
equal to 35 cm what comprises a minimum depth of the railway ballast and is the
most unfavourable case. It is important to include in the model of the backfill some
part of the space of the soil beyond the springings. It gives better representation of
this continuous medium behaviour. Boundary conditions for the backfill at the bottom
edges are fixed supports. Lateral edges of the backfill have only vertical movements
released what simulates continuity of this semi infinite medium beyond these edges.
4.4.4 Loads
Definition and application of loads comprises another crucial issue for analysis of
masonry bridge model. There are generally two different kinds of loads which has to
be distinguished: dead and live loads. Dead load comprise self-weights of each struc-
tural component according to its density. In numerical examples live load is assumed
to be represented by single axle load P of a vehicle standing on the structure. This
axle is acting on the whole width of the model what is obvious in case of two-
dimensional geometry model. In respect of presence of the truck structure the force
is dispersed in longitudinal direction. Therefore the live load is modelled as a uni-
formly distributed load on a length of 50 cm (Figure 4.7).
Another significant issue is the way and order of the loads application. Due to the
model nonlinearity and the loading history dependence all loads must be applied in-
crementally. Therefore in the proposed analysis during the first step the dead loads
are increased up to the final value and only then in the next step the live load starts
to act.
P, u
rigid 50 cm
plate
tained in the loading process described in the previous chapter. The value is deter-
mined by the highest point at the force-displacement relation presented in Figure 4.8
(light green line). In the presented numerical examples usually several centimetres of
vertical displacement provides the desired pick value of the load. The same proce-
dure is conducted first for undamaged and then for damaged structure. Such ap-
proach enables evaluation of a defect influence on the structure load capacity as a
measurable numerical change in the maximum load carried by it.
Pult
Pdult
Figure 4.8 Axle load-displacement relation (P-u) with marked ultimate load value Pult
A significant assumption made in this report is that the changes of the ultimate load
are proportional to changes of the load carrying capacity. According to most of codes
and regulations the load carrying capacity is a equal of the ultimate load divided by a
coefficient of safety. Taking this coefficient as a constant value a ratio between the
ultimate load of damaged and undamaged structures is equal to the ratio of the cor-
responding load carrying capacities. The former relationship is proposed to be ana-
lysed directly as a measure for the influence of defects on the load carrying capacity.
Hence a defect impact factor given on a percentage basis is defined:
where Pult is a difference of the ultimate loads for undamaged Pult and damaged
Pdult structure of the same geometry and material properties.
Percentage values of are given in the diagrams presenting results of the numerical
examples of modelling defects.
For each defect various locations of the live load are analysed and the critical one,
corresponding to the minimum value of Pult, is taken into account. Therefore relying
on symmetry of the structure only defects located in a half of the arch comprise vari-
ous cases. Cases from the other half give repeating solutions.
span: L0 = 5 m,
rise to pan ratio: r/L0 = 1/2, 1/4, 1/6,
arch barrel thickness: d = 45 cm,
height of the fill over crown: h = 35 cm,
Material parameters are the same as those given in chapter 4.4.
Result of the comparison are presented in Table 4.1 in a form of the ultimate value of
a single axle load set in a quarter point of a structure.
Table 4.1 Comparison the ultimate loads achieved by mans of various methods {kN]
r/L0 FEM RING* Archie-M
1/2 424.0 412 (406) 445
1/4 434.6 459 (398) 440
* - values in bracket for Boussinesq load dispersion
fcd,t
d = 100% eq. 18
fc,t
elements with
modified mate-
rial properties
a)
r/L0=0.25 m, Destr=0.3
[%]
b) 17
20 16
15 15
10
14
5
13
0
12
15
14 Ext=0.3
11
13
12
10
11
10 9
L0 [m]
9
0.5
8 0.45 8
0.4
7 0.35
0.3
0.25
6 0.2 Location 7
0.15
0.1
5 0.05
0
r/L0=0.25 m, Destr=0.3
20
[%]
c) 19
20
18
15
17
10
5 16
0
15
15 Ext=0.5
14
14
13
12
13
11
10
L0 [m]
9 12
0.5
8 0.45
0.4
7 0.35 11
0.3
0.25 Location
6 0.2
0.15
0.1
5 0.05
0
a)
r/L0=0.5 m, Destr=0.7
b) 18
[%]
20
16
15
14
10
5
12
0
15
Ext=0.3 10
14
13
12
8
11
10
L0 [m]
9
6
0.5
8 0.45
0.4
7 0.35
0.3
0.25
6 0.2 Location
0.15 4
0.1
5 0.05
0
r/L0=0.5 m, Destr=0.7
[%] c) 19
20
18
15
17
10
16
5
15
0
14
15
Ext=0.5
14
13
13
12
12
11
10 11
L0 [m]
9
0.5
8 0.45 10
0.4
7 0.35
0.3
0.25 Location
6 0.2 9
0.15
0.1
5 0.05
0
4.7.2 Discontinuity
B Bc B
c = 100% = 100% eq. 19
B B
where B, Bc and B are defined in Figure 4.1:
Figure 4.12 Top view of the arch barrel describing discontinuity parameters
plied by 1 meter. In this way it is assumed that there is no more interaction between
separated edge and the remaining central part of the arch barrel.
elements
with reduced
thickness
[%] 14
15
12
10
10
8
0
15
14
6
13
12
11 4
10
L0 [m]
9
0.7
8 2
0.6
0.5
7
0.4
6 0.3
0.2
Extent
5 0.1
0
[%]
14
15
12
10
5 10
0
8
15
14
13 6
12
11
4
10
L0 [m]
9
0.7
8
0.6 2
7 0.5
0.4
6 0.3
0.2
Extent
5 0.1
0
removed
elements
Figure 4.16 Modelling losses
a)
b)
a)
b)
= 0.125 a)
b)
= 0.250
c)
= 0.500
Figure 4.19 Results for loss of material, r/L0 = 1/4, a) = 0.125, b) = 0.250,
c) = 0.500
For r/L0 = 1/4 and = 0.40 (with = 10 %, = 20 %) the defect impact factor reaches
30 % and for r/L0 = 1/2 and = 0.50 the defect impact factor reaches 25 %.
Sustainable Bridges SB-4.7.3 2007-11-30 68 (71)
5 Final remarks
5.1 Conclusions
Commonly available methods for the assessment of masonry bridges provide possi-
bilities of simplified modelling of the main defect types. Selection of the appropriate
method is strongly dependent on the considered defect type. Some of the simple
methods are especially useful in the analysis of the typical defects providing suffi-
ciently precise results for comparison of the various defect cases. However some of
the defects require more advanced approach like those based of 3D FE models to
evaluate reliably their influence on the load carrying capacity of the structure.
Detailed analysis of the influence on the ultimate load of the selected types of defects
with their the most common parameter values is presented. It shows that such typical
defects can reduce the ultimate load of the structure by up to 30 % what is a signifi-
cant value indicating the need for consideration of this phenomenon. Also an impor-
tant finding is the most crucial locations for the defects. For all the analysed defect
types and structure geometries the critical locations are equal usually to about =
0.40 which coincides with location of one of the forming plastic hinges.
References
Ali S., Page A. W. (1988): Finite element model for masonry subjected to concen-
trated loads. Journal of Structural Engineering, 114, No. 8, pp.1761-1784.
Azevedo J., Sincraian G. (2001): Modelling The Seismic Behaviour Of Monumental
Masonry Structures. International Millennium Congress - Archi 2000, Paris.
Brencich A., De Francesco U. (2004a): Assessment of Multi-Span Masonry Arch
Bridges. Part I: a Simplified Approach, J. of Bridge Eng.ng, ASCE, 9, pp. 582-
590.
Brencich A., De Francesco U. (2004b): Assessment of Multi-Span Masonry Arch
Bridges. Part II: Examples and Applications. J. of Bridge Eng.ng, ASCE, 9, pp.
591-598.
Brencich A., De Francesco U., Gambarotta L. (2001): Non linear elasto-plastic col-
lapse analysis of multi-span masonry arch bridges. Arch'01, Paris, 19-21 Sep-
tember, pp. 513-522.
Choo, B.S., Coutie, M.G. and Gong, N.G. (1991a): Finite element analysis of ma-
sonry arch bridges using tapered beam elements. Proc. Inst. Civ. Engs. Part 2,
91, pp. 755-770.
Choo, B.S., Coutie, M.G. and Gong, N.G. (1991b): The effects of cracks on the be-
haviour of masonry arches. Proc. 9th International Brick/Block Masonry Confer-
ence, Vol. 2, pp. 948-955, Berlin.
Crisfield M. A. (1985): Finite element and mechanism methods for the analysis of
masonry and brickwork arches. Transport and Road Research Laboratory, Lon-
don.
Department of Transport (2001): The Assessment of Highway Bridges and Struc-
tures, Design Manual for Roads and Bridges: BA 16/97. Volume 3. Section 4.
Part 4.
Fanning, P.J., Boothby, T.E. (2001): Three-dimensional modelling and full-scale test-
ing of stone arch bridges. Computers & Structures, Vol. 79, pp. 2645-2662.
Fanning, P.J., Boothby, T.E., Roberts, B.J. (2001): Longitudinal and transverse ef-
fects in masonry arch assessment. Construction and Building Materials, Vol. 15,
pp. 51-60.
Ford T.E., Augarde C.E., Tuxford S.S. (2003): Modelling masonry arch bridges using
commercial finite element software. 9th International Conference on Civil and
Structural Engineering Computing, Egmond aan Zee, The Netherlands.
Frunzio G., Monaco M., Gesualdo A. (2001): 3D F.E.M. analysis of a Roman arch
bridge. Historical Constructions, P.B. Loureno, P. Roca (Eds.), Guimares, pp.
591-599.
Gilbert, M. (1998): On the analysis of multi-ring brickwork arch bridges. Proc. 2nd
International Arch Bridges Conference, Venice, pp. 109-118.
Gilbert, M. (2001): RING home page, http://www.shef.ac.uk/ring
Sustainable Bridges SB-4.7.3 2007-11-30 70 (71)
PRIORITY 6
SUSTAINABLE DEVELOPMENT
GLOBAL CHANGE & ECOSYSTEMS
INTEGRATED PROJECT
Sustainable Bridges SB-4.7.4 2007-11-30 2 (80)
This report is one of the deliverables from the Integrated Research Project Sustainable Bridges - Assessment for
Future Traffic Demands and Longer Lives funded by the European Commission within 6th Framework Pro-
gramme. The Project aims to help European railways to meet increasing transportation demands, which can only
be accommodated on the existing railway network by allowing the passage of heavier freight trains and faster
passenger trains. This requires that the existing bridges within the network have to be upgraded without causing
unnecessary disruption to the carriage of goods and passengers, and without compromising the safety and econ-
omy of the railways.
A consortium, consisting of 32 partners drawn from railway bridge owners, consultants, contractors, research
institutes and universities, has carried out the Project, which has a gross budget of more than 10 million Euros.
The European Commission has provided substantial funding, with the balancing funding has been coming from
the Project partners. Skanska Sverige AB has provided the overall co-ordination of the Project, whilst Lule Tech-
nical University has undertaken the scientific leadership.
The Project has developed improved procedures and methods for inspection, testing, monitoring and condition
assessment, of railway bridges. Furthermore, it has developed advanced methodologies for assessing the safe
carrying capacity of bridges and better engineering solutions for repair and strengthening of bridges that are found
to be in need of attention.
The authors of this report have used their best endeavours to ensure that the information presented here is of the
highest quality. However, no liability can be accepted by the authors for any loss caused by its use.
Table of Contents
1 Introduction ..........................................................................................................................4
2 Objectives ............................................................................................................................5
3 Failure modes and Limit State Functions ............................................................................6
3.1 Single ring arches.......................................................................................................6
3.2 Multi-ring arches.........................................................................................................7
3.2.1 Four hinge mechanism...................................................................................8
3.2.2 Ring separation ..............................................................................................8
4 Levels of assessment: failure criteria and theoretical models .............................................9
4.1 Axial- Bending failure .................................................................................................9
4.1.1 Monotonic loading ..........................................................................................9
4.1.2 Cyclic loading (fatigue) .................................................................................13
4.2 Ring separation ........................................................................................................13
4.3 Theoretical models ...................................................................................................14
4.3.1 Probabilistic approach ..................................................................................15
4.3.2 Simplified numerical models.........................................................................17
5 Basic variables...................................................................................................................54
5.1 Dimensions...............................................................................................................54
5.2 Materials...................................................................................................................54
5.3 S-N diagrams. Probabilistic fatigue models..............................................................54
5.3.1 S-N diagrams for masonry in compression-bending ....................................54
5.3.2 S-N diagrams for masonry in shear..............................................................65
6 Examples ...........................................................................................................................68
6.1 Assessment of ULS by 4-hinges mechanism ...........................................................68
6.1.1 Description of the bridge ..............................................................................68
6.1.2 Assessment assuming linear-elastic behaviour ...........................................70
6.1.3 Assessment based on non-linear response and system failure analysis .....71
6.2 Assessment of Fatigue LS .......................................................................................75
7 Conclusions .......................................................................................................................78
8 References ........................................................................................................................79
Sustainable Bridges SB-4.7.4 2007-11-30 4 (80)
1 Introduction
Several methods are available for the assessment of capacity of masonry arch bridges, from
very simple to more sophisticated, as presented in [1]. In this document, only the methods
based on the use of probabilistic procedures are presented to investigate their potentiality of
application in the assessment of masonry arches.
Techniques using reliability-based methods have been successfully applied to the assess-
ment of concrete and steel bridges. Nowadays, a lot of examples present in the available
technical literature show the advantages of the use of reliability-based approaches to the
assessment of existing bridges. However their application to the masonry bridges has been
almost negligible. One of the reasons is the difficulty to define reliable failure criteria for this
type of structures. Another reason is the lack of statistical data on material properties of ma-
sonry and filling material. The experimental tests show that normally the failure of an arch is
of a global nature more than due to the failure of a bridge component. In most cases, even
the division of the bridge in different components is almost impossible. For this reason, also
the lack of accurate theoretical models for the idealization of the behaviour of the masonry
arch bridge as a system, including the interaction effects with the filling material, spandrel
walls, etc., has been another difficulty not yet overcome. Last, but not least, the absence of
reliable data on the statistical definition of the material properties has been another issue that
has limited the use of probability-based assessment techniques. Very few data is available
for the behaviour of the masonry under static loads. The lack of experimental data is even
more dramatic in the case of cyclic loading.
However, some experiences have shown the potentiality of the use of reliability-based as-
sessment together with non-linear models of the bridge behaviour in the capacity assess-
ment of masonry arch bridges. In some cases, masonry bridges that will be rated as unsafe
when using standard methods of assessment and linear models, have reported high enough
level of safety when appropriate theoretical models and probabilistic assessment tools are
applied [2,3,4]. The global failure of the arch due to the formation of a hinged mechanism has
been the only failure mode considered so far. However, other response mechanisms (ring
separation, material crushing,) may lead the masonry arch to the failure and will be also
investigated in the present work. Finally, the serviceability issues related to the fatigue be-
haviour under service loads will be also considered, as they appear also to play an important
role in the service life of existing masonry arches [1]. As a consequence, a probabilistic ap-
proach to the fatigue failure assessment of masonry arch bridges is also presented in this
document.
Sustainable Bridges SB-4.7.4 2007-11-30 5 (80)
2 Objectives
Based on the existing achievements obtained till now in the use of reliability-based methods
for the assessment of existing bridges, the main objective is to investigate the potentiality of
probabilistic methods in the assessment of masonry arches, leading to an increase of the
load-capacity of these structures by comparison with the results from deterministic analyses.
To this end, the following objectives are considered:
1.- Definition of failure modes and limit state functions for masonry arch bridges
2.- Definition of mechanical models for accurate simulation of bridge behaviour up-to failure,
according to the identified failure modes in 1
3.- Derivation of simplified models for the reliability-based assessment of masonry arch
bridges
4.- Statistical definition of resistance random variables involved in the limit state functions.
Sustainable Bridges SB-4.7.4 2007-11-30 6 (80)
Figure 2.- Single ring arch formed by several units across the arch barrel depth
Despite being less feasible, the fatigue failure has to be also considered in this type of
arches, mainly when deterioration due to aging and service loads may derive on lower fa-
tigue strength of the masonry mix. The fatigue failure is due to repeated cycles of compres-
sion causing the crushing and squeezing of the mortar from the mortar joints followed by
vertical splitting of the bricks.
Ring separation
N 2M
G=H (1)
fc B N
Figure 6.- Collapse of the Prestwood bridge due to 4 hinges mechanism [5]
To take into consideration the behaviour up to failure of the bridge, more sophisticated theo-
retical models are necessary that take into account both material and geometrical non-
linearities. These models are able to simulate the formation of the hinges in the critical sec-
tions, as seen in figures 7 and 8 [4]. The bridge in figure 7 is a multi-span masonry-brick arch
bridge and the bridge in figure 8 is a single-span masonry- stone arch bridge. In both cases
the load is applied at the quarter span.
In figure 7, the black zones affecting the overall cross-section depth indicate the formation of
a hinge.
Sustainable Bridges SB-4.7.4 2007-11-30 11 (80)
(a)
b)
Figure 7.- Magarola arch bridge (Spain). General view (a) and stresses at failure (b).
Sustainable Bridges SB-4.7.4 2007-11-30 12 (80)
(a)
(b)
Figure 8.- Jerte arch bridge (Spain). General aspect (a) and stresses at failure (b)
In figure 8, the hinges are located in the cross-section with a full cross-section depth col-
oured in red.
The application of probability-based assessment methods when using such sophisticated
and advanced models is cumbersome. Because of the non-linear nature of the problem, an
explicit limit state equation is not available. The wisest approach is to use a simulation proc-
ess for the calculation of the probability of failure. However, crude Monte-Carlo simulation
can be hardly applied due to the huge number of runs necessary in the non-linear model and
the complexity of such a model. For this reason more advanced simulation techniques as the
importance-sampling, directional sampling and Latin Hypercube Method or other alternative
methods as the Surface Response method [6] are the most adequate tools to combine with a
full non-linear model. In chapter 6.1.3 is presented a practical application using the Latin Hy-
percube Method.
Simplified models to assess the ultimate capacity taking into account the non-linear behav-
iour are available through their calibration from the results of a full rigid block and plastic
Sustainable Bridges SB-4.7.4 2007-11-30 13 (80)
analysis. This is the case of the model developed by Martn-Caro et al. 2004 [7] . Explicit
equations are proposed for the maximum point and uniform distributed loads in terms of sim-
ple geometrical parameters and material strength. The model is of application in the case of
not skewed and not curved bridges, not multi-ring masonry bridges with consecutive regular
spans. The abutments should be un-deformable and the spandrels are solid. The parametric
study comprises spans in the range from 2 to 20 m., with a minimum rise to span ration of
1/6 and vault depth at crown to span ratio limited to a value between 0.05 and 0.10 depend-
ing on the span-length. The spandrel high at crown is in the range from 0.15 to 2 m. and the
maximum pier height is 10 m. Despite looking very restrictive, these requirements are ac-
complished by an important number of masonry bridges in the network.
The use of simplified models allows in some cases an explicit equation for the bridge capac-
ity, as a function of the geometric and material characteristics. In this way the load and resis-
tance parts of the problem become uncoupled and a Limit State Function in the form G=R-S
can be formulated. The reliability index can be then evaluated using FOSM or FORM.
A complete summary of the available numerical and analytical models to analyze the four-
hinge failure mechanism is reported in [8].
the two shapes shown in figure 9. For a low number of cycles with very high load, the arch
fails by the 4-hinge mechanism, up to a certain number of cycles of load (N*) where the criti-
cal mechanism becomes the ring separation.
Figure 9.- S-N curves for the coupled mechanism-ring separation fatigue behaviour
However, we should mention that this approach based on the definition of S-N curves for the
whole arch derives on an inapplicable methodology. In fact, the shape of the S-N curve in the
case of the bridge (system level) will depend on the specific characteristics of the bridge, with
its particular configuration of geometry, boundary conditions and loading pattern. Depending
on those parameters (among others), the fatigue effects due to the hinged mechanism or the
ring separation will be more or less enhanced. In fact, the level of the normal or shear
stresses governing both mechanisms depend on the mentioned variables. As a conse-
quence, the S-N curve obtained in a fatigue test will be only representative of the specific
conditions of such test, and will be hardly adopted in the case of a bridge with other condi-
tions. For this reason, the approach based on the definition of S-N curves at a specimen or
material level and not at the arch level will be adopted as presented in the next chapter. In
this case, the derivation of the S-N curves become also simpler as the fatigue tests may be
carried in small specimens and not in the whole arch structure.
From them, the first one should be analysed as an ULS as described in 4.1, but the other
three can be analysed from a material point of view involving a Fatigue or Serviceability LS at
service loads, avoiding in this way the problems deriving from the use of S-N curves at the
system level as mentioned in chapter 4.2
In fact, ring separation and sliding of masonry is provoked by the failure of the shear strength
in the interface between masonry and mortar. Crushing is controlled by the compressive
strength of masonry. In both cases, failure may appear due to a repeated number of cycles
of relatively low level of load. Depending on the specific characteristics of the bridge (geome-
try, boundary conditions, loading,.) the variable external loads will produce higher or lower
levels of compression and shear stresses in the material. The level of stress in each case
and the corresponding number of load cycles jointly with the fatigue strength will finally de-
cide if the first mechanism of failure to appear is the crushing of masonry or the sliding or ring
separation.
Therefore, the following methodology is proposed to model the behaviour of the failure
mechanism of ring separation, sliding and crushing when they are caused by accumulation of
cycles of load at service level (fatigue effect):
1.- Divide the problem in two analysis: one related to compression and the other to shear
interface stress in the masonry
2.- Use the Miners rule ( accumulated damage) as criteria for failure for both the compres-
sive strength and shear strength of the masonry
3.- Obtain S-N curves for both mechanisms of fatigue failure ( compression and shear)
4.- Identify in the bridge the more prone locations to suffer from excessive compression and
shear stresses
5.- Using a numerical model of the bridge, obtain the number of cycles and the stress incre-
ment (for both compression and shear) in the identified locations due to the external variable
loads. The rain-flow method can be used to this end.
6.- The bridge will fail when either the accumulated damage of Miner in compression or in
shear will reach a value equal to 1.0
ni
G= 1 (2)
i Ni
ni = number of cycles of load level i due to external loads (random variable)
Ni = number of cycles of load level i that the bridge can support (random variable)
ni as a random variable will be obtained via structural analysis taking into account the ran-
domness in the live-loads acting on the bridge or, alternatively, by extrapolation of measure-
ments taken in the bridge. Also the variability of the bridge properties should be considered
in the definition of ni as the resultant stress increment will also depend on this properties.
The way to derive the statistics of this random variable is by simulation of traffic effects jointly
Sustainable Bridges SB-4.7.4 2007-11-30 16 (80)
with simulation of geometric and material properties of the structure [11,12]. In the case of
railway traffic and taking into account that the stress increments due to bridge dynamics will
be very low for this type of bridges ( the infill mitigates the vibration level), one may consider
the passage of each convoy as a cycle of loading.
The process of simulation to derive the random variable ni requires the use of a theoretical
model where the input variables (geometry and mechanical properties of masonry) can be
modified according to their randomness. Several advanced FEM Codes are available for the
analysis of masonry arches. However in order to make simulation of practical application, a
numerical model as simple as possible is essential. The possibility of the use of simplified
models of analysis of masonry arches is discussed in chapter 4.3.2
The statistical definition of Ni can be done in the following way. Different works have shown
that the Weibull distribution function agrees very well with the expected physical criteria of
progressive fatigue deterioration [13,14]. On the basis of physically valid assumption, sound
experimental verification, relative ease in its use and better developed statistics, the Weibull
distribution has been widely used for the fatigue analysis of metals. It is also well-suited for
certain procedures of statistical extrapolation of large systems [13]. In [14] and [15], the
distribution of fatigue life of concrete was found also to approximately follow the Weibull
distribution. Some theoretical and experimental works [16,17] have shown also the feasibility
of the Weibull distribution regarding the statistical model for steel wires and strands to
fatigue. The randomness of the masonry' fatigue strength is also represented here by means
of a Weibull distribution. The reason for this assumption is fully demonstrated in chapter 5.3
The expressions for the Probability Density Function (PDF), fN(n), and the Cumulative
Distribution Function (CDF), FN(n), of a Weibull random variable are:
1
n n0 nn
f N ( n) = exp 0
, n n0 (3)
u n0 u n0 u n 0
nn
FN (n) =1 exp 0
, n n0 (4)
u n0
being = shape parameter, u = characteristic extreme value and n0 = minimum value. For
fatigue analysis, n0 is usually taken equal to 0.0 [13,14].
The first two moments of the variable are:
1
E ( N ) =u1 + (5)
2 1
var( N ) = u 2 1 + 2 1 + (6)
is the gamma function. For a complete definition of the fatigue strength in probabilistic
terms, the parameters should be known for all possible stress levels. This is not the case in
most situations because of lack of experimental data. In chapter 5.3, is explained how a
Weibull distribution for the fatigue strength of masonry and for several stress levels can be
obtained based on the few experimental data available at the present time.
Sustainable Bridges SB-4.7.4 2007-11-30 17 (80)
1045
8300
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
plan view
215
305
1665
750
395
8300
elevation
north abutment
south abutment
8300
1045
3000
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
plan view
215
305
2165
1665
750
395
8300
elevation
Class A Engineering bricks described as Nori bricks were used in the construction of the
arch barrel. The bricks had the following mean properties: compressive strength 154 N/mm2
and density 2370 kg/m3 and the nominal dimensions 215x102x65 mm. The mortar 1:2:9
(cement:lime:sand) mix by volume was used throughout the arch barrel. The mean compres-
sive strength was 1.9 N/mm2 as determined from five 100 mm cubes cured under the same
conditions as the arch barrel.
The first test bridge was filled with MOT Type 1 graded crushed limestone. The limestone
was compacted in 11 layers and the required weight of fill for each layer was compacted to
the required specification to control its density. An average density of 1.91 Mg/m3 was
achieved.
The second test bridge was filled with clay up to the level of the crown and compacted in 100
mm layers. The clay was described as a firm reddish brown slightly sandy CLAY with occa-
sional gravel with an optimum moisture content of 9%, liquid limit 29% and plastic limit 12%.
During filling, small samples were taken at regular intervals for moisture content testing. The
average moisture content was 13.4% based on 95 samples (standard deviation 1.4%); the
average density was found to be 2.21 Mg/m3 based on 14 tube samples taken from locations
across the clay body after the test (standard deviation 0.3 Mg/m3). The filling was finished
with two layers of limestone on the top of the clay.
The load was positioned vertically above the quarter point of the arch barrels
In order to study the influence of abutment fixity, two reinforced concrete abutments on which
the arch barrel were built were fixed 3000 mm apart, parallel to each other. Each abutment
comprised two parts; a lower section that was bolted to the structural strong floor, and an
upper section that could slide should the forces be large enough to overcome the
bond/friction between the blocks. The details of the two abutments are shown in Figure 12.
Sustainable Bridges SB-4.7.4 2007-11-30 19 (80)
154
230
215
13
530
530
20
395
322 325
Figure 12.- Construction details of abutments in limestone and clay filled bridge tests
Both bridges ultimately collapsed in four-hinged mechanisms (see figure 13). Whilst at col-
lapse hinged mechanisms formed, some abutment movement was recorded prior to this
stage. In the case of the crushed limestone filled bridge spreading of the abutment remote
from the applied load peaked at 3 mm, whilst in the case of the clay filled bridge this was
significantly greater, peaking at 8 mm. However, in the case of the clay filled bridge rotation
about the base of the lower abutment block remote from the applied load was identified, de-
spite the fact that this had been bolted to the structural strong floor. In the case of each
bridge, movement of the abutment closest to the applied loads was small < 1 mm.
Figure 14 shows the quarter span vertical deflections of both bridges tested. A peak load of
125 kN for the limestone filled arch and 85 kN for the clay filled, were obtained.
140
100
Bridge 2 (clay filled)
Applied load (kN)
80
60
40
20
0
0 5 10 15 20 25 30
Radial displacement (mm)
Figure 14.- Load- vertical displacement (quarter span) relationship for both tests
f c =0.6 f b
0.65 0.25
fm (7)
a value of fc = 18 MPa is obtained. Using the empirical relation Ec = 1000 fc, a value of 18000
MPa is obtained for the elasticity modulus of the masonry. Note that mean and non charac-
teristic values are used for material properties as the objective of the analysis is to check the
model with the results of the test.
A non-linear analysis was done for different values of the elasticity modulus of the infill mate-
rial, ranging from 3.5 to 3.5x102 MPa. The compressive strength of the infill material has
been also changed proportionally to the Young modulus. Once the self weight and perma-
nent loads are applied, the point load at the quarter point is incremented step by step at a
rate of 12.5 kN per step. In each step the program identifies the cracked elements. A maxi-
mum of 250 iterations was fixed. Close to the ultimate load, the applied load is adapted and
limited. The program verifies that the displacements corresponding to the applied load are
tolerable, i.e. less than the imposed ones. If this is not the case, the applied load is adapted
accordingly. The absence of convergence can be interpreted as the collapse of the arch.
The summary of the results for the ultimate load and the vertical displacement of the arch
below the point of load application ( quarter point) and the horizontal displacement in the
abutment more far from the load are presented in figure 16.
35
Horiz deplacement (mm)
30
Vert deplacement (mm)
Ultimate load (*10kN) 25
20
15
10
0
3,50E+06 1,80E+07 3,50E+07 1,80E+08 3,50E+08
Infill Young modulus
Figure 16.- Ultimate load, deflection at quarter span and horizontal displacement at
abutment as function of the infill Young modulus (in Pa)
From the figure, we can see that the value of the Young modulus that better identifies the
ultimate load for the case of the limestone infill, is 3.5x107 Pa, whereas a value of 1.2x107 Pa
is the most appropriate in the case of the clay infill.
In the figures 17 to 21 are shown the principal stresses in the structure for the different val-
ues of the elasticity modulus analysed.
Sustainable Bridges SB-4.7.4 2007-11-30 22 (80)
Finally, in figure 22 is shown the horizontal pressure on the arch for the different values of
the Young modulus of the infill material.
Sustainable Bridges SB-4.7.4 2007-11-30 24 (80)
140,00
70,00
60,00
50,00
40,00
30,00
20,00
10,00
0,00
-2,0E+05 -1,5E+05 -1,0E+05 -5,0E+04 0,0E+00 5,0E+04 1,0E+05 1,5E+05 2,0E+05 2,5E+05 3,0E+05
Pressure
Figure 22.- Variation of the horizontal pressure at failure on the arch barrel as a function of
the elasticity modulus of the infill material. The left side corresponds to the side where the
load is applied
From these figures and the evolution of the ultimate load and displacements with the rigidity
of the infill material ( figure 16), the following conclusions can be derived for the influence of
the arch-soil interaction:
1.- In a first stage, the vertical and horizontal displacements increase in an important way,
afterwards there is a decrease and finally there is a large increase for higher ultimate loads.
Therefore, the bridge performs quite differently depending on the stiffness of the backfill ma-
terial.
2.- For all values of the stiffness of the backfill, the failure is due to a 4 hinge mechanism.
However, as the stiffness increases, the hinges become more spread across the arch. This
can be due to the stress redistribution inside the backfill depending on the relative stiffness. 3
different behaviours can be observed:
a) For a very soft backfill, its collaboration to the resistance of the arch is negligible and the
behaviour is similar to that of an isolated arch barrel. The backfill does not transmit any load
to the soil and the arch and abutment remain fixed (no relative horizontal displacement be-
tween arch and abutment)
b) For a stiffer backfill, this gives and important resistance, enough for a relative displace-
ment between arch and abutment to appear.
Sustainable Bridges SB-4.7.4 2007-11-30 25 (80)
c) Finally, for a very stiff backfill, the backfill does not allow the horizontal displacement of the
abutment, transmitting a high load to the soil and spreading the external load to a higher
length of the arch.
3.- The horizontal pressure in the barrel is higher in the right side, i.e., the side where the
load is not applied. According to the 3 modes of behaviour just identified (see 2), the in-
crease in the stiffness of the infill material leads to the increment of the horizontal pressure in
the arch barrel and to a spreading of the applied load to an important part of the arch. This
derives in a higher resistance of the structure.
4.- The influence of the backfill material is critical to the strength of the masonry arch. De-
pending on the stiffness one identifies 3 different behaviours: arch isolated (the contribution
of the horizontal pressure of the backfill to the resistance is negligible), arch with collaborat-
ing backfill (horizontal pressure from backfill to the arch barrel increases the arch resistance)
and arch+backfill acting as a unique structure.
5.- The horizontal displacement in the abutment increases due to the common displacement
of the arch and the abutment, up to a level stiffness where a relative displacement between
arch and abutment (sliding force is higher that the frictional force in the joint) appears and,
consequently, the displacement decreases.
Simplified model
A simple theoretical model has been calibrated with the results of the tests and the advanced
model (DIANA software). The simplified model (ARCC) is based on the classical limit analy-
sis with and static approach, and the Lower Bound and Unicity theorems. However, the origi-
nal model takes into account the compression strength (no assumption of infinite compres-
sive strength of the masonry). The numerical model is very simple and implemented in a Ex-
cel sheet. In a first step, the model without arch-infill interaction has been used to the analy-
sis of the test, showing a bad prediction of the ultimate load. Afterwards, the interaction be-
tween arch and backfill has been included in the model what has led to a better accurate
prediction of the ultimate load.
When the influence of the infill material is not considered in the arch resistance, an ultimate
load of 43 kN is obtained. This is clearly much lower than the load predicted by the advanced
model and the test result (125 kN).
The interaction of arch and infill material has been considered in the model by using the re-
sults from the advanced numerical model (DIANA). The horizontal pressure over the arch
barrel corresponding to a stiffness of 3.5x107 Pa (limestone infill) as appears in figure 23 has
been used to calculate the horizontal force transmitted from the infill to the barrel and intro-
duced in the corresponding equilibrium of forces over the arch barrel. Taking into account
this interaction, the ultimate load is 122 kN, in good agreement with the advanced model and
the test for the case of limestone infill.
Sustainable Bridges SB-4.7.4 2007-11-30 26 (80)
140
130
120
110
3.5E7,Right Side
3.5E7,Left Side
100
Abutment
90
80
Height
70
60
50
40
30
20
10
0
-1,0E+05 -8,0E+04 -6,0E+04 -4,0E+04 -2,0E+04 0,0E+00 2,0E+04 4,0E+04 6,0E+04 8,0E+04 1,0E+05
Horizontal Pressure (Pa)
Figure 23.- Horizontal pressure (in Pa) over the arch extrados
A similar calculation has been carried out for the other values of the stiffness of the backfill
and the corresponding horizontal pressure curves shown in figure 22. The comparison of the
results obtained with the simplified ARCC model considering the barrel-backfill interaction
and the advanced model (DIANA) is presented in table 1 and figures 24 and 25.
Infill E modulus
DIANA Ultimate Load (kN) ARCC Ultimate Load (kN)
(Pa)
3,50E+08 287,5 182
1,80E+08 200 147
3,50E+07 125 122
1,80E+07 110 119
3,50E+06 55 72
Table 1.- Comparison of results from advanced (DIANA) and simplified (ARCC) theoretical
models as a function of Young modulus of infill material
Sustainable Bridges SB-4.7.4 2007-11-30 27 (80)
350
300
Ultimate load (kN)
250
200 Diana
150 Excel
100
50
0
0,00E+00 1,00E+08 2,00E+08 3,00E+08 4,00E+08
Infill Young's modulus (Pa)
Figure 24.- Comparison of results from advanced model and simple model ARCC (Excel)
190
170
Ultimate load (kN)
150
130 Diana
Excel
110
90
70
50
0,00E+00 2,00E+07 4,00E+07 6,00E+07 8,00E+07 1,00E+08
Infill Young's modulus (Pa)
Figure 25.- Comparison of results from advanced model and simple model ARCC (Excel)
As can be seen, a good approximation is obtained with the simple model if the interaction is
considered, for values of the Young modulus of the backfill material up to 35 MPa. For higher
values, the results from the simple model are not as accurate.
The ARCC model has been used to study the influence of the variation of the most significa-
tive parameters on the ultimate load for the case of a Young modulus of the infill equal to 35
MPa. The results are shown in figures 26 to 29.
Sustainable Bridges SB-4.7.4 2007-11-30 28 (80)
140
135
Ultimate load (kN
130
125
120
115
110
105
-35 -15 5 25
Variation of the infill density
(%)
Figure 26.- Influence of the density of the infill material on the ultimate load
210
190
Ultimate load (kN
170
150
130
110
90
70
50
-22 -12 -2 8 18
Variation of the arch barrel thickness
(%)
129
Ultimate load (kN
127
125
123
121
119
117
115
-11 -6 -1 4 9
Variation of the span-length
(%)
240
220
Ultimate load (kN
200
180
160
140
120
-16 -11 -6 -1
Variation of the rise
(%)
The models corresponds to the bridge 3-2 tested by Gilbert and Melbourne [18]. The bridge
geometry is presented in figure 30.
Figure 30.- Main dimensions of the bridge 3-2 tested by Gilbert and Melbourne [18]
The span to raise ratio is 4:1. The load was applied at the quarter of span. The two rings are
separated by damp sand rather than mortar to simulate the defect of ring separation and the
spandrel walls were detached in order to simulate also this defect in the bridge. The materi-
als used were Class A Engineering bricks in conjunction with a weak 1:2:9 (ce-
ment:lime:sand) mortar. The backfilling was made by graded crushed limestone compacted
in layers. The mean densities of the brickwork and backfill were measured as 2310 Kg/m 3
and 2260 Kg/m 3 . The bridge was loaded at quarter span. The failure load was 360 kN.
Figure 32.- Principal stresses for increasing level of the applied load
As shown in the example, DIANA accurately predicts the ultimate load where the ring sepa-
ration takes place and the stress levels due to service loads. Also the model is able to predict
Sustainable Bridges SB-4.7.4 2007-11-30 34 (80)
the load level where the ring separation will take place according to the load increase. How-
ever, from the works developed so far, it seems that simplified models as the Ring software
do not predict as accurate estimates for the ultimate load. In this case, the ultimate load re-
sulting from the RING software is 320 kN, something lower than the real 360 kN (figure 33).
Moreover, in the RING method, the assumption of ring separation has to be introduced in the
analysis from the beginning and is not obtained from the loading application process.
Therefore, at this stage and for multi-ring arches, it will be better to concentrate on the possi-
bility of using simplified models only to predict the stresses in the arch due to service loads.
First of all, the possibility of a linear elastic model should be explored, comparing their results
with those coming from more advanced models as DIANA.
Several tests have shown that a very influencing parameter in the load capacity of masonry
arch bridges is the skewnes at the supports. The main objective of the study herein pre-
sented is to investigate such influence and also to see the capability of existing theoretical
models to simulate the real behaviour of skewed arch bridges. Two theoretical models are
investigated. The first is based on a 3D solid FEM using software DIANA and the other is
based on a generalized advanced beam element using the software CRIPTA [4]
To validate the capacity of the FEM model in the analyses of skewed masonry arches, the
models tested by Melbourne and Hodgson [22] with angles of 0, 22.5 and 45 without span-
drel walls were selected (Table 3). These bridges spanned 3 m. The arches were segmental
with a maximum rise at midspan of 0.75 m and its depth was constant of 0.22 m. All arches
present the same supported length of 2.84 m (Figures 34 and 35).
Figure 34.- Geometry of bridges with 0 and 22.5 skew angles (m).
Loads acting on the bridge were its dead weight, including the spandrels infill and an exter-
nal line load (knife) at L/4 applied parallel to the abutments. The latter load was increased up
to reaching the failure. The spreading of the external loads through the infill was simulated by
a triangular distribution with an angle of 1H:2V. Boundary conditions at the abutments were
supposed perfectly rigid, therefore, all movements (degrees of freedom) were blocked.
A mesh of 3D solid (brick) elements were used for the arch: a ten nodes isoparametric tet-
raedric element was selected (Figure 36).
Models contained 12000 elements with a resulting amount of 20000 nodes (Figures 37 and
38). Dimensions of the mesh were determined by the depth of the arch and the length and
wideness of the arch. A critical aspect of the mesh was selecting the number of finite ele-
ments across the depth. That number has to be enough to allow huge gradients of deforma-
tion where hinges had to appear and rotate. However, total number of nodes (and degrees
of freedom) had to be controlled to have calculations within a reasonable time. After some
parametric studies, the depth of the arch was modelled by five elements and, thus, eleven
nodes. All models have eleven nodes in its depth and have varying number of nodes in longi-
tudinal and transverse direction (see table 4).
MATERIAL MODELLING
Masonry has been suposed to present isotropic mechanical properties and smeared crack-
ing. Past experiencies demonstrate that such approach is enough accurate when looking for
global plastic mechanisms in arches.
A total strain crack model has been selected, based on the Modified compression field theory
proposed by Vecchio and Collins, including the extension to 3D develped by Selby and Vec-
chio. The Total Strain crack model evaluates stresses in the directions defined by cracks.
Exponential softening was selected for cracking in tension. Figure 39 shows the stress strain
relationship after cracking in tension.
Sustainable Bridges SB-4.7.4 2007-11-30 38 (80)
cr
c
Factor is:
x = x =1 x = x =1 c
= y ( x)dx = y ( x)dx + 0dx = (1 x c )dx =
x =0 x =0 x =1 x =0 1+ c
G If
cr
nn .ult = 4.226
hf t
In compression, a perfect elastoplastic behaviour was choosen. In spite of being quite simple
it is a very good approach and improves the speed of calculations (Figure 41).
Sustainable Bridges SB-4.7.4 2007-11-30 39 (80)
Table 5 shows the material properties used in the analysis. High value of ultimate tension
strain was required to allow the formation of a full failure mechanisms.
Material properties
Deformation modulus E (kN/m2) 9000 Experimental
Poisson modulus 0.2 Experimental
Tensile strength ft (N/mm2) 0.05 Experimental
Compressive strength fc (N/mm2) 28.0 Experimental
Fracture energy Gf (Nm/m2) 100
Ultimate tension strain u 0.01
Table 5.- Material properties used in the analysis.
Full Newton-Raphson algorithm was used to solve the nonlinear problem. Convergence tol-
erances were 0.01 in relative value for both nodal displacements and nodal forces.
RESULTS
Results are presented for each skew.
0 skew
Figure 42.- Vertical displacements of the arch subjecte to dead load (m).
Sustainable Bridges SB-4.7.4 2007-11-30 40 (80)
Figure 43.- Deformed shape of the arch subjecte to dead load (x6000).
Figure 44.- Normal stresses under DL. in N/m2. Compressive stresses are negative.
The ultimate load achieved was 213 KN with a collapsing mechanism of four hinges (figures
45 and 46). Compression stresses are less than the compressive strength of the masonry
(28 N/mm2).
Figure 47 shows the evolution of the vertical displacement at quarter-span under the increas-
ing live load.
SKEW 0
250
200
Applied load (KN)
150
100
50
0
0 0,2 0,4 0,6 0,8 1 1,2
Displacement (mm)
Figure 47.- Load vs. displacement under the live load diagram.
22.5 skew
Under dead loads, compressive stresses are concentrated at the obtuse abutments. The
peak value is -0.55 N/mm2 (Figure 48). Figure 49 shows the distribution of horizontal force at
the springings.
Sustainable Bridges SB-4.7.4 2007-11-30 42 (80)
Figure 49.- Forces in X axis at the springings of the arch subjected to DL.
Maximum vertical displacement under DL is 0.06 mm (Figure 50) at the free sides of the
structure.
Sustainable Bridges SB-4.7.4 2007-11-30 43 (80)
Figure 51.- Stress state of the intrados and the extrados of the arch in N/m2
at 40%,60% and 80% of the ultimate live load.
Ultimate live load was 202 KN (5% less than the 0 skew arch). Figure 52 shows the evolu-
tion of the vertical displacement under the increasing live load.
Sustainable Bridges SB-4.7.4 2007-11-30 45 (80)
SKEW 22.5
250
150
100
50
0
0 0,2 0,4 0,6 0,8 1 1,2
Vertical displacement (mm)
Figure 52.- Load vs. displacement under the live load diagram.
The overall plan distribution of vertical displacements obtained by FE model of the bridge
with 22.5 skew presents good agreement with those registered in the test (Figure 53). Its
values are not directly comparable because only the dead weight of the infill was accounted
for in the FE model and not the horizontal pressure
45 skew
Under DL, this bridge presents larger vertical displacements and compressive stresses (Fig-
ure 54). Table 6 shows a comparison of such results obtained in the three models analyzed.
Figure 55.- Stress state of the intrados of the arch in N/m2 at different values of the ultimate
live load.
Figure 55 shows the evolution of the stress state of the intrados of the bridge subjected to an
axle live load. It is worth noting that the formation of the hinge at the springing more distant
from the axle load is more complex than in the other bridges. Figure 56 present the cracks
and damages at the extrados reported by Melbourne with the stress state at failure obtained
in this work. There is a good agreement between maximum compressive stresses and
cracked zones of the numerical model with those from the test.
Figures 57 and 58 illustrate the inhomogeneity of stresses and cracking at different cross
sections parallel to the springings and parallel to the arch sides, respectivelly.
Figure 58.- Stress state at failure in different cross sections parallel to the arch side.
The overall plan distribution of vertical displacements obtained by FE model of the bridge
with 45 skew presents good agreement with those registered in the test (Figure 59). The
values are not comparable because only the dead weight of the infill was accounted for in the
FE model. Ultimate load obtained by the model was 104 KN and the experimental 337 KN.
Sustainable Bridges SB-4.7.4 2007-11-30 49 (80)
ULTIMATE LOAD
600
500
400
Axle load (KN)
300
200
100
0
0 10 20 30 40 50 60
Skew angle (in )
Experimental results [22] and numerical results of Table 7 and Figure 60 show that the load
capacity of skewed bridges with an angle from 0 to 22.5 is almost the same of the right one.
The sophisticated beam model was unable to yield accurate results when increasing the
skew. Table 7 and Figure 60 show that ultimate loads are systematically underestimated
when compared with those of the experimental tests. That is because the constraints intro-
duced by the infill to the movements of the arch are neglected, despite taking into account
the weight and the spreading of the live load were accounted for the infill. However, tenden-
cies shown by experimental tests are very well reflected in the analysis of skewed arches
without simulating the stiffness and strength of the infill. In fact, there is proportionality be-
tween experimental and numerical results varying the skew (Table 8)
SKEW 22.5
CRIPTA DIANA
1800
1600
1400
Failure load (KN)
1200
1000
800
600
400
200
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 61.- Diagram Load vs. Position on the arch with a 22.5 skew.
SKEW 45
CRIPTA DIANA
1600
1400
1200
Failure load (KN)
1000
800
600
400
200
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 62.- Diagram Load vs. Position on the arch with a 45 skew.
Sustainable Bridges SB-4.7.4 2007-11-30 52 (80)
SKEW 45
DIANA
300
250
Failure load (KN)
200
150
100
50
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
Figure 63.- Diagram Load vs. Position on the arch with a 45 skew.
As a reference, detailed analysis of the worst position for the axle load on the right bridge is
presented in Figure 64. The results were obtained by using the sophisticated beam model
(CRIPTA).
SKEW 0
CRIPTA
2000
1800
1600
Failure load (KN)
1400
1200
1000
800
600
400
200
0
0 0,1 0,2 0,3 0,4 0,5 0,6
x/L
All results show the high load capacity of loading these arches at mid-span. Such result was
well known in the case of right bridges, where a five hinge mechanism is required when sub-
jected to this symmetric loading. The analysis performed show that skewed bridges have the
same behaviour when subjected to axle load applied at mid-span.
CONCLUSIONS
Experimental results [22] and numerical results herein presented show that the load
capacity of skewed bridges with an angle from 0 to 22.5 when failing by the 4-hinge
mechanism is almost the same of the right one. However, compression stresses in
the springings are concentrated at the obtuse angle area in the case of skewed
bridges.
Failure mechanisms of arches presenting and skew over 22.5 are fully 3D.
From the examples analysed can be concluded that, in the case of maintaining con-
stant the distance between supports, varying skew angle do not change the worst po-
sition of the axle load on the platform. So, the worst position of the load can be de-
duced in a simple model.
As it is well known, the effect of the infill constraining the movements of the arch af-
fect significantly the load capacity. In the numerical examples presented, only the
weight and the spreading of the live load were accounted for the infill. So that, load
capacities were systematically underestimated. However, tendencies shown by ex-
perimental tests are very well reflected in the analysis of skewed arches without simu-
lating the stiffness and strength of the infill. In fact, there is a clear proportionality be-
tween experimental and numerical results varying the skew.
With the models used, 3D nonlinear finite element calculations for skew arches were
highly time-consuming.
As could be foreseen, sophisticated beam elements were not capable to fully simu-
late the ultimate capacity of masonry arch bridges with large skew, despite simulating
complex formation of hinges. Results on bridges with a skew between 0 and 22.5
present very good agreement with those provided by 3D solid FE.
Sustainable Bridges SB-4.7.4 2007-11-30 54 (80)
5 Basic variables
For a probabilistic approach to the assessment of masonry arches, the statistical definition of
the basic variables is mandatory. The statistical definition of the basic variables should be
based on the available data existing in the literature up-dated with the measurements ob-
tained directly on the bridge under assessment. The up-dating can be carried out using
Bayesian techniques with the data available from the inspection and monitoring in the bridge.
The most important data that needs to be collected in order to improve the quality of the pre-
dicted carrying capacity using probabilistic techniques is shown in the following.
5.1 Dimensions
The most representative dimensions for a probabilistic analysis of masonry arches are the
arch thickness, arch width, depth of fill at crown, rise at mid-span and span-length. At
this time, there is not enough data available to obtain the statistical definition of these vari-
ables and in the literature information how to get the statistical parameters is not available
too. Based on some measurements made by the authors in several masonry arch bridges, a
rough estimate of 10 % coefficient of variation for the arch thickness and 5 % for the arch
width can be assumed. The distribution function can be assumed as Normal. Because these
values have been obtained from a very limited number of samples, it is recommended that
up-dating of this information should be done from results gathered in the specific bridge un-
der assessment.
5.2 Materials
The most representative material properties for a probabilistic assessment of a masonry arch
are the compressive strength of masonry, tensile strength of masonry, compressive
strength and modulus of elasticity of filling material, , density of masonry and filling
material, endurance limit to fatigue of masonry and shear strength (interface masonry-
mortar). At the present time there is not available data in the literature on the statistical dis-
tribution and parameters of these properties.
m
Concrete: N c max =K (8)
f cm
Steel: N m = K (9)
where cmax is the maximum applied stress, fcm is the compressive strength of the concrete,
N is the number of cycles to failure, is the applied stress increment, m and K are the char-
acteristic power and constant associated to a specific branch of the S-N curve. In equation
(8) it is assumed that the minimum stress level is zero and therefore the stress range is equal
to cmax
Assuming that fatigue phenomena causing crushing of mortar and brick subjected to com-
pression is similar to that of concrete, Hwans methodology could be adopted here to define
also the fatigue strength of masonry. Assuming a Weibull distribution, and noting that s is a
constant deviation of log N for a given log , for all stress increments, the parameters of the
distribution could be estimated through the following expressions [11]:
2
2 = (10)
6s 2
0.5772
ln u = + ln( K m ) (11)
The values K and m can be obtained from the available fatigue tests on masonry samples in
compression. Normally, a linear regression through experimental data reported in a log N-
log graph is used.
However, the direct application of Hwans proposal to masonry structures is highly debatable
mainly for two reasons:
1.- The fatigue behaviour of the material may be sensible not only to the stress range applied
to the material, but also to the minimum level of stress imposed during the cycle loading.
2.- It has to be demonstrated in some way that the assumption of a constant value of s (what
according to equation (10) implies a constant value for the parameter ) independent from
the stress level is valid.
The only experimental data in masonry specimens tested to fatigue in different conditions is
available in [9]. They conclude that the high cycle fatigue strength of wet and submerged
brick masonry test specimens is only slightly less than the fatigue strength of similar labora-
tory dry test specimens and propose the following equation for the lower bound fatigue
strength:
where F(S) is the so-called function of induced stresses. S and Smax are the induced stress
range and maximum induced stress respectively and Su is the quasi-static compressive
strength under similar loading conditions and N is the number of constant amplitude load
cycles. As shown in equation (12), the fatigue strength depends not only on the stress range
but also on the maximum and minimum level of stress in the material.
Sustainable Bridges SB-4.7.4 2007-11-30 56 (80)
In order to obtain the parameters of the statistical distribution of the number of cycles to fail-
ure, one has to perform a statistical analysis of fatigue data at every stress level S. It is very
difficult to do this directly when the stress level and the stress range are both variables. For
this reason, the variable F(S) as defined in equation (12) was first adopted in [9] as indicative
of the stress level in the material.
According to the expression of the Weibull distribution with two parameters ( equation (4)
with n0 = 0)
n
FN (n) =1 exp , n 0 (13)
u
This equation can be used to verify if the experimental data obtained from a series of tests
follows the Weibull disitribution. From the experimental data, the survival probabilities SF for
each fatigue life N can be calculated according to SF = 1-[m/(k+1)] where k = total number of
N data for a certain value of the stress level FS, and m = the ascending order of the N data
point considered. By means of a linear regression to the experimental values, the parameters
and u can be calculated.
From the data available in [9], the figures 65 to 69 were obtained. They draw the regression
analysis for different values of the function of induced stresses, F(S).
Sustainable Bridges SB-4.7.4 2007-11-30 57 (80)
FS=0.70
1
y = 1,0565x - 9,668
R2 = 0,6873
0,5
0
0 2 4 6 8 10 12
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
FS=0.65
1
y = 0,2666x - 3,2458
2
R = 0,8255
0,5
0
0 2 4 6 8 10 12 14 16
-0,5
ln(ln(1/SF)
Serie1
Lineal (Serie1)
-1
-1,5
-2
-2,5
ln(N)
FS=0.6
1
y = 0,3524x - 4,4408
R2 = 0,8089
0,5
0
0 2 4 6 8 10 12 14 16
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
FS=0,55
0,5
y = 0,4386x - 6,3641
2
R = 0,8936
0
0 2 4 6 8 10 12 14 16
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
FS=0.50
0
0 2 4 6 8 10 12 14 16
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(N)
As can be observed from the figures, the obtained correlations are not very good. For this
reason, according to the results reported in [15], where they propose a probabilistic fatigue
model for plain concrete, the following fatigue formula is proposed to take into account the
minimum value of repeated stress:
max
log S = log =1 b(1 R ) log N (16)
u
S is the ratio of the maximum loading stress to the strength and R is the ratio of the minimum
stress to the maximum stress min / max. In order to work with only one stress variable in-
stead of two, the introduction of the equivalent fatigue life EN is done. EN is defined as
EN=N(1-R). In this way, equation (16) takes the form of the Wholer equation (logS= a-b log
EN).
In the following figures are presented the results of the linear regression of the experimental
data from [9] with the new variables S and EN instead of FS and N
Sustainable Bridges SB-4.7.4 2007-11-30 60 (80)
S= 0.9
0,5
y = 0,8511x - 5,336
R2 = 0,9603
0
0 1 2 3 4 5 6 7 8
ln(ln(1/SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(EN)
S=0.8
y = 0,5353x - 4,3585
R2 = 0,9772
0,5
0
0 1 2 3 4 5 6 7 8 9 10
ln(ln(1(SF))
Serie1
-0,5
Lineal (Serie1)
-1
-1,5
-2
ln(EN)
S= 0.75
y = 1,0753x - 7,7298
2
0,5 R = 0,9746
0
0 1 2 3 4 5 6 7 8 9
-0,5
ln(ln(1/SF))
Serie1
Lineal (Serie1)
-1
-1,5
-2
-2,5
ln(EN)
S=0.70
y = 0,4604x - 4,8822
2
R = 0,8753
0,5
0
0 2 4 6 8 10 12 14
-0,5
ln(ln(1/SF))
Serie1
Lineal (Serie1)
-1
-1,5
-2
-2,5
ln(EN)
S=0,65
0,6
0,4
y = 0,2379x - 2,521
0,2
R2 = 0,5046
0
0 2 4 6 8 10 12 14
-0,2
ln(ln(1/SF))
Serie1
-0,4
Lineal (Serie1)
-0,6
-0,8
-1
-1,2
-1,4
ln(EN)
S=0.60
0,6
y = 0,4202x - 5,3679
0,4 2
R = 0,9703
0,2
0
0 2 4 6 8 10 12 14 16
-0,2
ln(ln(1/SF))
Serie1
-0,4
Lineal (Serie1)
-0,6
-0,8
-1
-1,2
-1,4
ln(EN)
S=0.55
0,6
0,4
0,2
y = 0,8785x - 9,0308
R2 = 0,8778
0
8,6 8,8 9 9,2 9,4 9,6 9,8 10 10,2 10,4 10,6
-0,2
ln(ln(1/SF))
Serie1
-0,4
Lineal (Serie1)
-0,6
-0,8
-1
-1,2
-1,4
ln(EN)
As seen in the previous figures, the correlation is much better than in the previous case.
Therefore, equation 16 is proposed in the present study as the most appropriate S-N curve
for masonry subjected to compression. From the regression analysis, the corresponding pa-
rameters of the Weibull distribution for different values of S are presented in table 9.
S lnu u r2 r
0.9 0.8511 5.336 528 0.96 0.98
0.8 0.5353 4.3585 3436 0.98 0.99
0.75 1.0753 7.7298 1324 0.97 0.98
0.70 0.4604 4.8822 40306 0.88 0.94
0.65 0.2379 2.521 40010 0.50 0.71
0.60 0.4202 5.3679 353144 0.97 0.98
0.55 0.8785 9.0308 29138 0.88 0.94
Table 9.- Parameters and u of Weibull distribution for different values of stress level S
As can be seen from table 9, the regression coefficients for all stress levels are very high
except for the case S = 0.65. Also, from table 9, it becomes clear that the hypothesis of con-
stant value of deviation s ( or constant value of ) for different stress levels is not valid.
The parameters and u in table 9, depending on the stress level, should be used for the
reliability-based assessment to fatigue of existing masonry arches. In fact, as Roberts et al.
Sustainable Bridges SB-4.7.4 2007-11-30 64 (80)
point out in their paper [9], the low compressive strength of the brick and mortar used to
manufacture the test specimens were intended to be representative of the type of bricks and
mortar likely to be encountered in relatively old masonry arch bridges.
Depending on the number of cycles of load N for each stress level S present in the loading
histogram of the bridge, either the reliability index to fatigue failure and/or the remaining ser-
vice life with a certain confidence level can be obtained.
S = A x N-B(1-R) (17)
the values in table 10 are obtained for the coefficients A and B for different values of the sur-
vival function SF
SF A B r
0.95 1.106 0.0998 0.95
0.90 1.303 0.1109 0.98
0.80 1.458 0.1095 0.97
0.70 1.494 0.1023 0.93
0.60 1.487 0.0945 0.90
0.50 1.464 0.0874 0.86
Table 10.- Parameters of fatigue equation depending on the required confidence level
All data
0
0 1 2 3 4 5 6 7
-0,1
-0,2
Dry
wet
-0,3 submerged
SF=0.95
log S
SF=0.7
-0,4 Run out
Lineal (SF=0.95)
Lineal (SF=0.7)
-0,5
-0,6
-0,7
log (EN)
Figure 77.- Experimental data (all tests) and proposed fatigue equation for probability of fail-
ure of 5 and 30 %.
SN curve
Anticipated ISN curves
100
Mec hanism Ring separation
Mechanism
Load (% of static load)
Ri Ring
ng s epar
se at ion
pa
rat
i on
10
5m 'strong' arch
3m 'strong'arch
5m 'weak' arch
1
1
10
100
1,000
10,000
100,000
1,000,000
10,000,000
Number of cycles
Figure 78.- Results of laboratory tests for high cycle loading [1]
Sustainable Bridges SB-4.7.4 2007-11-30 66 (80)
Figure 78 is a summary of the results obtained in the Salfords tests. As shown there, the
fatigue mechanism of ring-separation can also be represented by an straight line in a log S
versus log N diagram. The diagram in figure 78 was obtained from tests carried out in the
whole arch and not in small specimens. For this reason, despite the materials for the 5 and 3
m strong arches are the same, a different behaviour in fatigue is observed depending on the
archs span. The main difference between both arches is the span-length and, as a conse-
quence, the stress level in the masonry for the case of self-weight plus permanent load
(minimum stress) and the stress level due to the applied test load (maximum stress). As dif-
ferent stress levels during the tests lead to different S-N diagrams, we may assume, as in the
case of the masonry subjected to compression, a stress versus number of cycles to failure
relationship in the form:
The two different straight lines in figure 78 for the strong brick and the 3 m and 5 m span can
be, then, explained by a different value of R (stress minimum/stress maximum) in the tests in
the 3 and 5 m span arches.
The results from the tests in Salford show and endurance limit around 40 % of the maximum
static load for 3 m span arches and good quality of brick. A similar endurance limit can be
defined in the case of 5 m span arches made of weak brick.
Therefore, based on the very few data available, it seems that a similar format of the S-N
diagram can be assumed for the masonry in compression and in shear. Only the coefficients
A and B will be different. As presented in the previous chapter, the values of A and B have
been derived in this work for fatigue of masonry in compression. To obtain these coefficients
in the case of shear, we propose a set of tests similar to those reported in [9], but where the
predominant action is the shear stress in the mortar between bricks. The idea is to test small
specimens and not complete arches as the last is much more difficult and expensive. Tests
on a representative number of specimens will allow to define if a Weibull distribution also fits
well the experimental results in this case, and to obtain the parameters of the distribution.
From this, the procedure of fatigue analysis as described in chapter 4.3, based on Miners
rule, can be applied.
Up to the moment when the results of the tests of masonry in shear may be available, the
proposal is to use some rough estimates of parameters A and B based on the fatigue tests in
full arches carried out at Salford University and described in [1]. According to equation 19
and the experimental values shown in figure 47, we can obtain the following relationships:
a) 3 m strong arch
b) 5 m strong arch
According to the data from the tests, we may roughly assume the values R1 = 0.42 and R2 =
0.56, what gives the following results: A= 2.69, B = 0.3 for the 3 m arch, and A = 1.20, B=
0.11 for the 5 m arch.
Taking into account the few experimental data available and until more experimental data is
obtained, the fatigue equation proposed is:
The methodology to apply for the assessment of multi-ring arches to fatigue will be similar to
that proposed for single-ring arches. In this last case, it is necessary to have models to de-
rive the compression in the most critical points of the arch as a function of time depending on
the external variable actions. In the same way, in the former case, the availability of models
to obtain the shear stresses in the contact between rings will be necessary to perform a ser-
vice life analysis. However, because the actions due to the passage of the normal trains will
not cause a high stress level, even a linear-elastic model can be used to calculate the com-
pression and shear stresses at defined critical points of the arch.
Depending on the span-length of the arch, the number of cycles of load to take into account
will be the total number of axles or the total number of trains passing the bridge. The vibra-
tion amplitudes due to dynamic effects of the train loads will be very small for this type of
bridges due to the damping through the infill, and, therefore, can be neglected in the calcula-
tion as they will be under the endurance limit.
Sustainable Bridges SB-4.7.4 2007-11-30 68 (80)
6 Examples
6.1 Assessment of ULS by 4-hinges mechanism
The example of application corresponds to the Magarola river bridge. The Magarola bridge is
a masonry arch bridge located near Barcelona in one of the busiest highways in Spain, con-
necting the city with Spains capital, Madrid. The structure is a multi-span arch bridge with 5
arches of 20 m each. A probabilistic capacity assessment is carried out, starting with the ex-
perimental testing program in order to get information about the geometric dimensions and
the strength characteristics of the materials, as no documentation is available at all. This is
normally the case of most existing masonry arches. The safety of the bridge is obtained in
terms of the probability of failure (reliability index). The safety level is evaluated for the actual
existing traffic which is available from the traffic surveys and using theoretical models with
different level of complexity and accuracy (from linear elastic at section level to non-linear at
system level).
Despite being a highway bridge, the methodology presented here is of full application to rail-
way bridges, only changing the characteristics of the live load.
The piers are built of rubble stone masonry reinforced with ashlars in their corners.
Because no drawings or other documentation was available, all the data used in the study
were obtained during the geometric and mechanical characterisation of the bridge carried out
in April 1997. The main geometric parameters of the arches and the mechanical properties of
the materials derived from the testing program are shown in Tables 11and 12 respectively.
The shape of the arches was defined by obtaining the co-ordinates of twenty-one (21) points.
This made possible to prove the exactitude of the semi-circumference they describe. Six ver-
tical boreholes of 2.10 m to 2.25 m (Figure 81) and four horizontal ones of 1.5 to 3.17 m
(Figure 81) allowed for recognition of the inside geometry of the bridge and its composition,
and obtaining several cores to be tested in the laboratory (Figure 82). Some of the tested
cores, which included brick and mortar joints, gave valuable information about the masonrys
strength and the characteristics of the filling material. Horizontal boreholes detected concrete
backing on the four piers.
Figure 82.- Column of the materials extracted from a vertical hole ( pave-
ment+concrete+backfill material+ masonry)
Sustainable Bridges SB-4.7.4 2007-11-30 70 (80)
Arch ring
Compressive (MPa) (experi- 15.0
Strength ment)
Deformation (MPa) (assumed) 3000
modulus
Tensile strength (MPa) (assumed) 0.01
Density (kg/m3) (experi- 1800
ment)
Fill
Compressive (MPa) (assumed) 1.0
Strength
Deformation (MPa) (assumed) 30
modulus
Density (kg/m3) (experi- 2300
ment)
Table 12.- Average material properties derived from the tests or assumed
variability as appear in table 13 were deduced. Also in the table are displayed the values
obtained for N and M at the most critical sections using a linear-elastic model
G= R-S (21)
R is the ultimate resistance of the bridge taken as a system, i.e., the load that the bridge can
carry up to the development of the failure mechanism. S is the external traffic action. In order
Sustainable Bridges SB-4.7.4 2007-11-30 72 (80)
to statistically characterize the random variable R, a simulation process was carried out using
a deterministic model for analysis of masonry structures up to failure [4]
Description of the non-linear numerical model
The numerical model performs the nonlinear analysis of masonry spatial structures consist-
ing of curved, three-dimensional members with variable cross section. The detailed deriva-
tion leading to the global equations of the structural problem, together with some details re-
garding its numerical implementation, may be found in the paper by Molins and Roca [4].
The application of the model to multi-span arch bridges is demonstrated in [19].
The global constitutive material model adopted results as a combination of partial constitutive
equations for masonry subjected to tension, compression and shear stresses. Interaction
between the normal and shear responses is considered through the use of a biaxial strength
envelope in order to define the limiting set of strength values.
The behaviour of the material subject to compressive stresses is described by means of a
bilinear elasto-plastic model. While in tension, masonry is modelled as a simple linear elastic
perfectly brittle material. Either cracking or separation between blocks occurs once the ten-
sion strength of the fabric is reached. No further softening behaviour is considered after the
first cracking. Shear behaviour is modelled by means of a bilinear elasto-plastic curve, similar
to that adopted for brick masonry under compression, with the limit strength max determined
from the strength envelope. In this case, a Mohr-Coulomb criterion, with no cohesion, was
adopted in order to model the shear response as a residual resisting mechanism due to fric-
tion between blocks and mortar in joints. The shear strength max is thus expressed as:
max = (22)
where is the applied compressive strength and is a coefficient of friction. Under tension,
the shear strength is made equal to zero.
As has been demonstrated in chapter 4.3.2. realistically evaluating the load capacity of barrel
vault bridges, such as Magarola, requires taking into account the horizontal action of the infill
and the spandrel walls, in addition to its dead load.
Figure 83.- Modelling of an arch bridge with the infill as a set of equivalent linear members.
Sustainable Bridges SB-4.7.4 2007-11-30 73 (80)
Non-linear analysis of the structure under dead loads showed that the arch rings are quite
uniformly compressed with the exception of a short zone at the extrados of the springings.
Sustainable Bridges SB-4.7.4 2007-11-30 74 (80)
Figure 85 shows the evolution of the load factor () for a five axle vehicle along the second
arch. This factor means the number that applied to the axle loads of the average 5-axle truck
(400 kN) produces numerical prediction of failure. In all positions failure is due to crushing in
compression at the crown of the second arch. That crushing occurs for a quasi collapsing
mechanism involving the formation of three hinges in the arch, one in each pier (not com-
pleted), three in the first arch of the bridge and two (not completed) in the third arch (adja-
cent to the second). Figure 86 shows the deformed shape of the bridge at failure. As can be
seen, the contour conditions induced by the flexibility of the piers and the adjacent arches
increases the flexibility of the loaded arch. They also permit detecting the huge rotation ex-
perienced by the sections around the crown of the second arch. In fact, the comparison of
these results with those obtained supposing rigid supports at the springings of the arch
shows that the loss of load capacity induced by the actual support conditions is around 50%.
Performing a simulation process based on the Uptaded Latin Hypercube Method [6] and us-
ing a Kolmogorov-Smirnof fitting technique, it was found that the load factor,, is well ap-
proximated by a Lognormal variable with mean equal to 56.7 and standard deviation 10.6.
The surveyed traffic data with the distribution of total weight in 5-axle trucks, identified S as a
Normal random variable with mean 2.0 (two 5 axle truck side by side) and a coefficient of
variation (COV) equal to 25 %. According to these values, a reliability index = 12.9 was
obtained for the actual traffic. As can be seen, the safety of the bridge using a reliability-
based methodology with a suitable and accurate model of the bridge behaviour up to failure
can be assured. Much higher reliability level is obtained compared to the linear analysis.
In conclusion, when a model that fully reflects the complex and non-linear behaviour of multi-
span arch bridges is used, the corresponding reliability indices are high enough to assure the
correct performance. Contrary, the assessment has also shown how assuming a simplified
and not as accurate model, the corresponding results may be completely unrealistic and lead
to inappropriate decisions concerning bridge safety. The non-linear model predicts a mecha-
nism of collapse involving adjacent piers and arches. The analyses carried out showed the
Sustainable Bridges SB-4.7.4 2007-11-30 75 (80)
importance of taking into account the complete structure instead of a single arch. In fact, the
analysis for single arches would have produced an overestimation of 100% of the load ca-
pacity. These two conclusions are very important regarding the future assessment of other
multi-span arch bridges
The assessment of the fatigue performance of a bridge can be formulated in two different
ways:
1.- Calculation of the reliability index or probability of failure in a given reference time
2.- Calculation of the remaining service life with a predefined probability level.
With the available experimental data, the first assessment will be only pertinent in the case of
fatigue due to compression in the masonry (single-ring arches). The process of the calcula-
tion of the reliability index or probability of failure due to service loads causing a fatigue fail-
ure is as follows:
1.- Build up a theoretical model of the bridge. Because the load level introduced in the analy-
sis will be equivalent to the service loads, a linear elastic model or other simplified models as
presented in this document can be used.
2.- Define the typical trains passing the bridge determining the axle-spacing and axle-loads.
Define how many of each typical trains will cross the bridge in the remaining service life pe-
riod (defined by the user)
3.- Simulate the passage of each train in the numerical model and obtain the stress versus
time relationship for the most critical locations in the bridge. Take into account that the loads
should include the dynamic amplification factor as we are dealing with service load levels.
Sustainable Bridges SB-4.7.4 2007-11-30 76 (80)
4.- From the stress-time relationship obtain the maximum stress in the predefined locations
and obtain the relation S= maximum stress/ compressive strength for each typical train and
only for the stress peaks in the stress-time curve where S > 0.5 (endurance limit)
5.- For each value of Si > 0.5 and according to the number of trains that will cross the bridge
in the defined period of time, obtain the number of times, ni , that each value S will be
reached in the selected locations ( Define the histogram of S ). According to what is ex-
plained in the next point, it is recommended to obtain the values of ni for the ranges
0.5<S<0.55, 0.55<S<0.6, 0.6<S< 0.65, 0.65<S<0.7, 0.7<S<0.75,0.75<S<0.8, 0.8<S.
6.- According to the Miners rule, the limit state function can be defined as:
ni
G=1
S i > 0.5 N i
(23)
where Ni is the number of loading events causing the failure. According to what is explained
in this document, Ni is a random variable that can be modelled by a Weibull distribution with
the parameters as presented in table 14 depending on the value of S. For each of the 7
ranges of S defined in point 5, the corresponding parameters of the distribution can be ob-
tained from table 14.
S lnu u
> 0.8 5.336 528
0.75-0.8 4.3585 3436
0.70-0.75 7.7298 1324
0.65-0.70 4.8822 40306
0.60-0.65 2.521 40010
0.55-0.60 5.3679 353144
0.5 - 0.55 9.0308 29138
Table 14.- Parameters of the Weibull distribution as a function of the stress level
7.- Once the random variables Ni present in the limit state function are perfectly defined, then
applying FORM, the reliability index can be calculated.
The process of the calculation of the remaining service life with a predefined probability level
is as follows:
1.- Proceed from step 1 to five as in the previous case
2.- Calculate the minimum stress level (compression or shear) in the selected locations. This
will be the value corresponding to the empty bridge (self-weight + permanent loads). For
each stress range, calculate the ration Ri = minimum stress/maximum stress
3.- In the case of fatigue due to compression (single-ring arches), according to equation 17
and the values of A and B in table 10 for the desired probability level, calculate for each
range of S (with the corresponding values of R) the number of events to failure Ni
Sustainable Bridges SB-4.7.4 2007-11-30 77 (80)
4.- In the case of fatigue due to shear (multi-ring arches), at this moment the only fatigue
equation available is equation (20), which corresponds to a probability of 50 % (correlated
directly to test data). Using this equation, the number of events to failure Ni can be estimated.
5.- Given the estimated number of typical trains crossing the bridge for a reference period ( 1
year for instance), then calculate the time necessary to reach the value 1 in the accumulated
damage equation.
Sustainable Bridges SB-4.7.4 2007-11-30 78 (80)
7 Conclusions
As used for other materials like concrete and steel, the potentiallity of probabilistic methods
in the assessment of masonry arches has been studied. The main problems encountered for
such an approach are mainly the response behaviour of this type of bridges that makes very
difficult an accurate modelization as well as the different construction methods used since
many years (single-ring, multi-ring,.). Many different materials as brick, mortar, infill mate-
rial with very different strength properties also introduce challenging problems to the model-
ling. Finally, and of special focus in the case of a probabilistic approach, is the lack of ex-
perimental data available in the literature necessary to derive a statistical definition of the
main properties of the materials. Since an accurate probabilistic assessment mainly relies on
an appropriate model of the structure to simulate the real behaviour and on an appropriate
statistical definition of the random variables involved, still an important work has to be devel-
oped in the coming years for a correct estimation of the safety of these bridges in terms of
their probability of failure or reliability index. However, with the still limited experimental data
available, some experiences have shown the potentiality of probabilistic methods applied to
masonry arches, resulting in final assessments that may leave the bridge as it is without
costly repair and/or strengthening remedial measures.
The present document is a first step in this sense, as it presents a complete methodology for
the probabilistic assessment of masonry arches at the serviceability and ultimate limit states.
The document explains the definition of the different failure modes and corresponding limit
state functions that may occur depending on the type of masonry construction (mainly single-
ring and multi-ring). Also the introduction of the possibility of the fatigue failure of masonry
arch bridges and the proposal of new assessment methods based at the serviceability level
[21] and not only at the ultimate level of the 4 hinge-mechanism is a promising initiative in the
field. As a consequence, the present document has deeply investigated the potentiality of a
probabilistic approach in the estimation of fatigue safety and remaining service life of ma-
sonry arch bridges. As a result, the need of simple numerical models to perform fast simula-
tion trials has been enhanced. Thus, the influence of important features in the bridge re-
sponse, as the arch-backfill interaction, the modelling of multi-ring arches and the bridge
skewness in the development of simplified theoretical models at the ultimate and serviceabil-
ity limit states have been worked out too.
The analysis developed in the present work has verified the results obtained in laboratory
tests regarding the influence of skewness in the failure load by 4-hinge mechanism. In fact, it
has been confirmed that the influence of a skew angle less than 22.5 is almost negligible.
The study has confirmed that only with advanced 3D FEM models is possible to correctly
model the behaviour up to failure of masonry arches with skew bigger than 22.5. It has been
also concluded that the worst load position for the arch can be deduced using a simple
model.
In summary, a methodology is presented for the fatigue and serviceability assessment of
masonry arches, and probabilistic S-N curves for the behaviour of masonry under fatigue are
developed based on the few experimental data available. In the case of masonry under com-
pression, a fatigue equation with various levels of probability of failure is also proposed that
may be used for deterministic assessments. In the case of masonry under shear a fatigue
equation with a 50 % confidence level is also proposed. The last can be of relevant applica-
tion to the case of multi-ring arches. Of course, these are only preliminary models and these
fatigue models have to be updated as more experimental data becomes available from future
laboratory and full-scale tests.
Sustainable Bridges SB-4.7.4 2007-11-30 79 (80)
8 References
[1] D4.7.1. Structural assessment of masonry arch bridges. Background document. Sustain-
able Bridges, 2007.
[2] Casas, J.R.: Assessment of masonry arch bridges. Experiences from recent case stud-
ies. Current and future trends in bridge design, construction and maintenance. Thomas Tel-
ford, 1999, pp. 575-586
[3] Casas, J.R.; Molins, C.: Assessment of the Magarola arch bridge. A case study. Pro-
ceedings of the Sixth nternational Conference on Structural Studies, Repairs and Mainte-
nance of Historical Buildings, Dresden, 1999, pp. 333-342
[4] Molins, C.; Roca, P.: Capacity of masonry arches and spatial structures. Journal of
Structural Engineering, ASCE, 124, (6), 1998, pp. 653-663
[5] Page, J.: Masonry arch bridges. State of the art review. Transport Research Laboratory,
UK, 1993
[6] D4.4.3. Probabilistic non-linear analysis. Background document. Sustainable Bridges
2007.
[7] Martn-Caro, J.A.; Martnez, J.L.; Len, J.: A first level structural analysis tool for the
Spanish railways masonry arch bridges. Proceedings of Arch Bridge IV, Advances in As-
sessment, Structural design and construction. CIMNE, Barcelona, 2004, pp.192-201
[8] D4.7.3. Methods of analysis of damaged masonry arch bridges. Background document.
Sustainable Bridges, 2007.
[9] Roberts, T.M.; Hughes, T.G.; Dandamudi, V.R.; Bell, B.: Quasi-static and high cycle fa-
tigue strength of brick masonry. Construction and Building Materials, Vol. 20, 2006, pp. 603-
614
[10] Melbourne, C.; Tomor, A.; Wang, J.: Cyclic load capacity and endurance limit of multi-
ring masonry arches. Proceedings of Arch Bridge IV, Advances in Assessment, Structural
design and construction. CIMNE, Barcelona, 2004, pp. 375-384
[11] Crespo-Minguillon, C.; Casas, J.R.: Fatigue reliability analysis of prestressed concrete
briges. Journal of Structural Engineering, ASCE, 124, (12), 1998, pp. 1458-1466
[12] Crespo, C. ; Casas, J.R., A comprehensive traffic load model for bridge safety checking,
Structural Safety, 19, (4), 1997, pp. 339-359.
[13] ASCE: "Foreword to Fatigue and Fracture Reliability. A state of the Art Review ". Jour-
nal of the Structural Division,ASCE, 108,1982, pp. 1-88.
[14] Hwan, B.: " Fatigue analysis of Plain concrete in Flexure ". Journal of Structural Engi-
neering, ASCE, 112, (2), 1986, pp. 273-289.
[15] Shi, X.P.; Fwa, T.F., Tan, S.A.: Flexural fatigue strength of plain concrete. ACI Materi-
als Journal, 90, (5), 1993, pp. 435-440
[16] Castillo, E. and Fernndez, A.: "Statistical models for fatigue analysis of long elements".
Proceedings of the IABSE Workshop on Length Effect on Fatigue of Wires and Strands, El
Paular, Madrid, 1992, pp. 15-32.
[17] Fernndez-Canteli, A., Castillo, E., Argelles, A.: "Length effect on fatigue of wires and
prestressing steels". Proceedings of the IABSE Workshop on Length Effect on Fatigue of
Wires and Strands, El Paular, Madrid, 1992, pp. 125-135.
Sustainable Bridges SB-4.7.4 2007-11-30 80 (80)
[18] Gilbert, M.; Melbourne, C.: Analysis of multi-ring brickwork arch bridges. Arch Bridges.
Proceedings of the First International Conference on Arch Bridges. Thomas Telford, 1995,
pp. 225-238
[19] Molins, C, Numerical simulation of the ultimate response of arch bridges. Chapter of the
book Structural analysis of historical constructions II, ed. Roca, Gonzlez, Oate and
Loureno, CIMNE, Barcelona, 1998, pp. 93-123
[20] Sicilia, C., Ultimate analysis of masonry bridges (in Spanish). Master Thesis. Universitat
Politcnica de Catalunya (UPC), Barcelona, 1997.
[21] Melbourne, C; Tomor, A. : A new assessment method for masonry arch bridges. Pro-
ceedings of IABMAS06. Taylor and Francis. Porto, 2006.
[22] Melbourne C., Hodgson J.A.: The behaviour of skewed brickwork arch bridges. First
International Conference on Arch Bridges, Thomas Telford. London, 1995, pp. 302-320.